Solar filament debris distribution method, apparatus, device, medium, and program product

Abstract

The application discloses a solar filament fragment distribution method and device, equipment, medium and program product, and relates to the technical field of astronomical information analysis. The method comprises the following steps: acquiring a to-be-processed full-disk H-alpha image; inputting the to-be-processed full-disk H-alpha image into a semantic segmentation model to obtain binary mask data of the to-be-processed full-disk H-alpha image; determining the connected domain distribution of the solar filament based on the binary mask data, taking the connected domain fragments of the solar filament as graph nodes, adding logical edges between any two nodes, and constructing an initialized full connection graph; realizing message transmission between nodes in the initialized connection graph based on a fragment node message transmission mechanism, updating node features and edge states, and obtaining an updated full connection graph; classifying the edges of the updated full connection graph, and based on the classification result, re-distributing the connected domain fragments of the solar filament to determine the filament attribution of each connected domain fragment. The application guarantees the complete tracking and continuity description of the filament in the life cycle.

Description

Technical Field

[0001] This application relates to the field of astronomical information analysis technology, and in particular to a method, apparatus, device, medium, and program product for allocating solar filament fragments. Background Technology

[0002] Solar filaments are an important phenomenon in solar activity, characterized by low temperature and high density. They typically appear as dark, strip-like structures suspended on the solar surface. Their eruptions are often closely related to solar flares and coronal mass ejections, significantly impacting the Earth's space environment and potentially causing severe interference to satellites, radio communications, and power systems. Studying the formation and evolution of solar filaments not only helps to deepen our understanding of the physical mechanisms of solar activity but also provides crucial clues for predicting solar activity, making it of great significance for space weather forecasting.

[0003] In related technologies, the study of solar filament fragment attribution typically employs two methods: a priori method and a data-driven method. The former usually involves first segmenting or labeling filaments from solar observation images using filament identification techniques, resulting in multiple fragment regions with varying shapes and spatial discontinuous distributions. Then, the distance and tilt angle between fragments are used to determine if they belong to the same filament. While computationally simple and efficient, this method suffers from limitations in universality due to the need for threshold selection for distance and tilt angle. Relying on manually set rules and thresholds, it often leads to misclassifications for cases outside the thresholds. The latter method, representatively, utilizes instance segmentation models from deep learning. During dataset creation, solar filament fragments are transformed into different instances of the same type; that is, without changing the category label, fragments are segmented based on their attribution relationships. However, despite the strong feature learning capabilities of instance segmentation models, simultaneously performing fragment attribution and segmentation tasks is limited by the pixel-level annotation quality of the data and the elongated, irregular, and complex morphology of the solar filament fragments themselves. This leads to difficulties in accurately characterizing fragment boundaries during prediction, resulting in overly smooth edges or misclassification of boundaries. This affects the accuracy of subsequent fragment attribution, thereby interfering with the reconstruction of the solar filament evolution process. Summary of the Invention

[0004] The main objective of this application is to provide a method, apparatus, device, storage medium, and computer product for allocating solar filament fragments, aiming to solve the technical problem of insufficient accuracy in allocating solar filament fragments in related technologies.

[0005] To achieve the above objectives, this application proposes a method for allocating solar filament fragments, the method comprising:

[0006] Obtain the full solar surface H-alpha image to be processed;

[0007] The full-solar H-alpha image to be processed is input into the semantic segmentation model to obtain the binary mask data of the full-solar H-alpha image to be processed;

[0008] Based on binary mask data, the distribution of connected components of solar filaments is determined. Using the connected component fragments of solar filaments as graph nodes, logical edges are added between any two nodes to construct an initial fully connected graph.

[0009] Based on the fragment node message passing mechanism, message passing between nodes in the initial connection graph is realized, node characteristics and edge states are updated, and the updated fully connected graph is obtained.

[0010] The edges of the updated fully connected graph are classified, and based on the classification results, the connected component fragments of the solar filament are redistributed.

[0011] In one embodiment, based on binary mask data, the distribution of connected components of solar filaments is determined. Using the connected component fragments of solar filaments as graph nodes, logical edges are added between any two nodes to construct an initial fully connected graph. The steps include:

[0012] For each connected component fragment in the binary mask data, a feature vector consistent with the location of the connected component fragment is determined in the high-level feature map output by the semantic segmentation model.

[0013] Spatial mapping and aggregation are performed on the feature vectors of connected component fragments to obtain high-dimensional representations of each connected component fragment.

[0014] Based on the multilayer perceptron, the high-dimensional representation after nonlinear transformation is used as the node features of the initial fully connected graph to construct the initial fully connected graph.

[0015] In one embodiment, the steps of initializing message passing between nodes in the connected graph, updating node features and edge states, and obtaining the updated fully connected graph based on the fragment node message passing mechanism include:

[0016] Centered on the edges of the initialized fully connected graph, for each edge, the state representation of each edge of the initialized fully connected graph is determined based on the node characteristics of the two endpoints of the edge;

[0017] Using the edges of the initialized fully connected graph as message passing channels, the nonlinear transformation node features of the source node are concatenated with the state representation of the corresponding message passing channel to obtain the message to be passed from the source node. Through the corresponding message passing channel, the message passing from the source node to the target node is realized, and the message passing of all nodes in the initialized fully connected graph is completed.

[0018] For each message passing channel, update the state representation of the message passing channel based on all messages flowing to the target node corresponding to the message passing channel;

[0019] For each node in the initialized fully connected graph, based on the updated channel state of the message transmission channel connected to the node, the node in the initialized fully connected graph is updated, completing one update cycle;

[0020] Based on a preset number of iterations, the initial fully connected graph is updated cyclically to obtain the updated fully connected graph.

[0021] In one embodiment, taking the edges of the initialized fully connected graph as the center, and for each edge, determining the state representation of each edge of the initialized fully connected graph based on the node features after nonlinear transformation of the nodes at both ends of the edge, includes:

[0022] For each edge, the node features after nonlinear transformation of the nodes at both ends of the edge are concatenated to obtain the node concatenation features.

[0023] After linearly transforming the node splicing features based on the initial edge weight parameters, a nonlinear transformation is performed using the hyperbolic tangent activation function to obtain the initial state of the edges.

[0024] In one embodiment, for each message passing channel, the step of updating the state representation of the message passing channel based on all messages flowing to the target node corresponding to the message passing channel includes:

[0025] For each message passing channel, the state representation of the message passing channel is updated based on the first gated loop unit, taking all messages flowing to the target node corresponding to the message passing channel and the current state representation of the message passing channel as input.

[0026] In one embodiment, the step of updating the node of the initialized fully connected graph based on the updated channel state of the message transmission channel connected to the node for each node initialized fully connected graph includes:

[0027] For each node in the initialized fully connected graph, the node in the initialized fully connected graph is updated based on the second gated loop unit, using the mean of the updated channel states of all message transmission channels connected to the node and the node characteristics of the initialized fully connected graph as input.

[0028] Secondly, to achieve the above objectives, this application further provides a solar filament debris distribution device, the device comprising:

[0029] The acquisition module is used to acquire the full solar surface H-alpha image to be processed;

[0030] The feature extraction module is used to input the full-solar H-alpha image to be processed into the semantic segmentation model to obtain the binary mask data of the full-solar H-alpha image to be processed;

[0031] The fully connected graph construction module is used to determine the connected component distribution of solar filaments based on binary mask data. It uses the connected component fragments of solar filaments as graph nodes and adds logical edges between any two nodes to construct an initial fully connected graph.

[0032] The node update module is used to implement message passing between nodes in the initial connection graph, update node features and edge states, and obtain the updated fully connected graph based on the fragment node message passing mechanism.

[0033] The redistribution module is used to classify the edges of the updated fully connected graph and redistribute the connected component fragments of the solar filament based on the classification results.

[0034] Thirdly, to achieve the above objectives, this application further provides a solar filament fragment allocation device, the device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the solar filament fragment allocation method as claimed in any one of claims 1 to 6.

[0035] Fourthly, to achieve the above objectives, this application further provides a storage medium, which is a computer-readable storage medium, and stores a computer program thereon. When the computer program is executed by a processor, it implements the steps of the above-described solar filament fragment allocation method.

[0036] Fifthly, to achieve the above objectives, this application further provides a computer program product, which includes a computer program that, when executed by a processor, implements the steps of the above-described solar filament fragment allocation method.

[0037] One or more technical solutions proposed in this application have at least the following technical effects:

[0038] This application transforms the problem of assigning solar filament fragments into a graph node redistribution problem, achieving a more refined, robust, and physically consistent automatic assignment of discontinuous solar filament fragments. Compared with geometric rule methods that rely on manual thresholds, this application does not require setting fixed thresholds such as distance and angle, and can adapt to morphological changes of filaments under different observation conditions, scales, and complex topologies. Furthermore, compared to traditional techniques that use instance segmentation models to simultaneously complete both fragment attribution and segmentation tasks, this application addresses fragment attribution and dark filament fragment region segmentation separately. This reduces the processing load on a single model, improves the accuracy of fragment segmentation, and significantly enhances the attribution accuracy of small fragments, elongated fragments, and fragments with blurred boundaries through high-dimensional feature extraction and a circular message flow mechanism at the edge center. By constructing connected component fragments into a fully connected graph and introducing bidirectional message propagation with structure flow awareness, this application can comprehensively consider the texture features, spatial structural relationships, and topological dependencies of fragments to achieve global modeling of the overall structure of solar dark filaments. This effectively reduces misassignment and missed associations, thereby improving the processing accuracy of fragment attribution. With improved accuracy in both fragment segmentation and fragment attribution, this application ensures complete tracking and continuous description of dark filaments throughout their lifecycle, providing a more reliable data foundation for subsequent solar activity evolution analysis and space weather forecasting. Attached Figure Description

[0039] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0040] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, those skilled in the art can obtain other drawings based on these drawings without creative effort.

[0041] Figure 1 This is a flowchart illustrating the solar filament fragment allocation method of this application.

[0042] Figure 2 This is a flowchart illustrating the cyclical update process.

[0043] Figure 3 This is a schematic diagram illustrating the effect of redistributing dark bars.

[0044] Figure 4 Schematic diagram of a solar filament debris distribution device.

[0045] Figure 5 Schematic diagram of equipment for distributing solar filament debris.

[0046] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0047] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application.

[0048] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0049] Based on the problems mentioned in the background art, this application provides a method for allocating solar filament fragments, referring to... Figure 1 , Figure 1 This is a flowchart illustrating the first embodiment of the solar filament fragment allocation method of this application.

[0050] In this embodiment, the design idea of ​​the solar fringe fragment allocation method is as follows: First, the fringe fragment regions distributed on the solar surface are determined by a semantic segmentation model; then, nodes are constructed for each fragment region based on the high-level features of semantic segmentation, and logical edges are connected between all fragments to establish an initial fully connected graph; next, a loop message passing is performed with the edges as the center, and the edge state and node state are updated by dual gating units to iteratively learn the structural consistency relationship between fragments; after the message fusion is stable, the edges are classified to determine whether the fragments should belong to the same fringe; finally, the fragments are reassigned according to the predicted edge connection relationship to generate a complete fringe region with continuous structure and physical consistency.

[0051] Specifically, in this embodiment, the solar filament fragment allocation method includes steps S10~S50:

[0052] Step S10: Obtain the H-alpha image of the entire solar surface to be processed.

[0053] Step S20: Input the H-alpha image of the full solar surface to be processed into the semantic segmentation model to identify solar dark stripes and obtain the binary mask data of the H-alpha of the full solar surface to be processed.

[0054] It should be noted that the H-alpha image of the entire solar surface to be processed refers to the solar observation image to be detected. The H-alpha band is sensitive to the structure of the solar chromosphere and can clearly reveal activity features such as dark streaks and flares. The binary mask data is the result of semantic segmentation through a semantic segmentation model, generally a binary image where pixel values ​​represent the segmentation result (e.g., 1 for dark streak regions, 0 for background). Based on this image, the connected component distribution of dark streaks can be determined for subsequent graph construction and fragment allocation. Connected components refer to independent regions formed by adjacent identical pixels in the image; their distribution describes the location and number of these regions, with each region representing an independent dark streak fragment.

[0055] It is understandable that solar filament fragments arise from the evolution of solar filaments due to various solar activities; therefore, a redistribution of these fragments is necessary to determine their affiliation to the filaments they belong to. The redistribution steps are as follows:

[0056] Step S30: Based on the binary mask data, determine the distribution of connected components of the solar filaments. Using the connected component fragments of the solar filaments as graph nodes, add logical edges between any two nodes to construct an initial fully connected graph.

[0057] Step S40: Based on the fragment node message passing mechanism, message passing between nodes in the initial connection graph is implemented, node features and edge states are updated, and the updated fully connected graph is obtained.

[0058] Step S50: Classify the edges of the updated fully connected graph, and based on the classification results, redistribute the connected component fragments of the solar filament.

[0059] Specifically, this embodiment first extracts all connected component fragments based on binary mask data and treats each fragment as a node in the graph. Then, logical edges are established between any two nodes to construct an initial fully connected graph that can describe the potential relationships between fragments. Subsequently, iterative information flow is carried out along each logical edge through a fragment node message passing mechanism, so that the node features and edge states are gradually updated in multiple rounds of loops. This allows the model to integrate the shape features, texture features, and spatial structure relationships of the fragments to learn whether they belong to the same solar filament. The updated logical edges are then classified to determine the affiliation relationship between nodes, and the connected component fragments are re-integrated based on the classification results to achieve continuous structural reconstruction and final affiliation assignment of the solar filament.

[0060] In one feasible implementation, step S30 includes steps A10 to A30:

[0061] Step A10: For each connected component fragment in the binary mask data, determine the feature vector that is consistent with the location of the connected component fragment in the high-level feature map output by the semantic segmentation model.

[0062] Step A20 involves spatial mapping and aggregation of the feature vectors of connected component fragments to obtain high-dimensional representations of each connected component fragment.

[0063] Step A30: Based on the multilayer perceptron, the high-dimensional representation after nonlinear transformation is used as the node features of the initial fully connected graph to construct the initial fully connected graph.

[0064] It should be noted that the high-level feature map is the layer convolution output of the semantic segmentation model, which can capture the abstract semantic information of the image. The feature vector extracted from it is used in this embodiment to represent the node features of the initialized fully connected graph.

[0065] Specifically, the steps of spatial mapping and aggregation of the feature vectors of connected component fragments to obtain high-dimensional representations of each connected component fragment can be expressed as follows:

[0066]

[0067] in, x i Indicates the first i A high-dimensional representation of solar filament fragments in feature space. R i Indicates the first i Within a region of solar dark filament fragments, ( u,v F( represents the pixel coordinates of the high-level feature map or binary mask data.) u,v ) represents the feature vector corresponding to the pixel coordinates in the high-level feature map.

[0068] Specifically, for each connected component in the mask i (Corresponds to a fragment of a sun's dark stripe) i ), based on the coordinates of all pixels within it ( u,v Find the feature vector that is in the same position in the high-level feature map. F ( u,v Then, for this connected component... i The feature vector F( at all internal locations) u,v After aggregation, the mean value is taken to obtain a high-dimensional representation of the fragment in the feature space, thus completing the spatial mapping and aggregation processing of the feature vectors of the connected domain fragments.

[0069] Subsequently, a multilayer perceptron (MLP) is used to perform a nonlinear transformation on the high-dimensional representations of each solar filament fragment, which serves as the node features for initializing each node of the fully connected graph. This can be represented as:

[0070]

[0071] in,x i Indicates the first i A high-dimensional representation of solar filament fragments (fully connected graph nodes) in the feature space. h i Indicates the first i Node characteristics of a fully connected graph.

[0072] Understandably, multilayer perceptrons can perform nonlinear reconstruction and feature enhancement of fragment features, making node features more discriminative and expressive. At the same time, they can compress noise and highlight the structural features that distinguish different fragments, thereby providing more stable and semantically rich node inputs for subsequent graph neural network-based relationship modeling, fragment attribution inference and edge feature message passing, significantly improving the accuracy and robustness of fragment association reasoning.

[0073] Furthermore, after constructing the initial fully connected graph, the fully connected graph is updated using a fragment node message passing mechanism. Specifically, in a feasible implementation, step S40 includes steps B10 to B50:

[0074] Step B10: Taking the edges of the initialized fully connected graph as the center, for each edge, determine the state representation of each edge of the initialized fully connected graph based on the node characteristics of the two endpoints of the edge.

[0075] Step B20: Using the edges of the initialized fully connected graph as message passing channels, the nonlinear transformation node features of the source node are concatenated with the state representation of the corresponding message passing channel to obtain the message to be passed from the source node. Through the corresponding message passing channel, message passing from the source node to the target node is realized, and message passing of all nodes in the initialized fully connected graph is completed.

[0076] Step B30: For each message passing channel, update the state representation of the message passing channel based on all messages flowing to the target node corresponding to the message passing channel.

[0077] Step B40: For each node in the initialized fully connected graph, update the node in the initialized fully connected graph based on the updated channel state of the message transmission channel connected to the node, and complete one update cycle.

[0078] Step B50: Based on a preset number of iterations, cyclically update the initial fully connected graph to obtain the updated fully connected graph.

[0079] Step B10 includes steps B11 to B12:

[0080] Step B11: For each edge, the node features after nonlinear transformation of the nodes at both ends of the edge are spliced ​​together to obtain the spliced ​​node features.

[0081] Step B12 involves linearly transforming the node splicing features based on the initial edge weight parameters, followed by nonlinear transformation using the hyperbolic tangent activation function to obtain the initial state of the edges.

[0082] Step B30 includes step B31:

[0083] Step B31: For each message passing channel, take all messages flowing to the target node corresponding to the message passing channel and the current state representation of the message passing channel as input, and update the state representation of the message passing channel based on the first gated loop unit.

[0084] Step B40 includes step B41:

[0085] Step B41: For each node in the initialized fully connected graph, using the average of the updated channel states of all message transmission channels connected to the node and the node characteristics of the initialized fully connected graph as input, the node in the initialized fully connected graph is updated based on the second gated loop unit.

[0086] Reference Figure 2 This implementation takes the edges of the initialized fully connected graph as the center, performs the cyclic flow of messages, establishes a cyclic structure, and presets a total of steps to achieve the orderly transmission of solar dark stripe nodes and edge features, and realizes the effective flow of solar dark stripe structure information.

[0087] Specifically, in the first loop of message passing, the edges of the initialized fully connected graph are used as message passing channels to establish the initial state representation of each message channel, which can be represented as:

[0088]

[0089] in, h i and h j Represents nodes in a fully connected graph i Node characteristics and nodes j Node characteristics, w edge_init Let represent the learnable edge weight matrix, whose initial values ​​are based on a standard normal distribution. b This represents the bias vector. e ij Represents a node i and nodes j The state representation of the message passing channel between them.

[0090] It can be seen that this implementation method incorporates node features h i and node features h jVector concatenation is performed to obtain [ i,j After that, through the initial edge weight matrix w edge_init The bias vector is mapped to a new feature space, and then nonlinearly activated using the hyperbolic tangent activation function to obtain the nodes. i and nodes j State representation of message passing channels between e ij .

[0091] Furthermore, based on the state representation of each message passing channel in the fully connected graph, the messages to be sent by each node are determined. These messages can be represented as:

[0092]

[0093] Where, m i←j Indicates from the source node j To the target node i The message to be sent (the message to be sent). h j Represents the target node j Node characteristics, w trans This represents the learnable message weight matrix. b This represents the channel bias vector. e ij This represents the state of the message transmission channel between target node i and source node j.

[0094] It is understandable that the source node is the sender of the message, and the target node is the receiver of the message. It can be seen that this implementation method represents the state of the message transmission channel between the target node i and the source node j, and the node characteristics of the source node... h j After vector concatenation, the message weight matrix is ​​used. w trans The bias vector is mapped to a new feature space, and then activated by the linear rectification function (ReLU) to obtain the message to be transmitted from the source node j to the target node i.

[0095] It is understandable that each node in a fully connected graph can act as either a source node or a target node. When acting as a target node, after the information transmission between all source nodes connected to the current target node, the messages transmitted by other source nodes are aggregated to form the node characteristics of the current node, which can be identified as follows:

[0096]

[0097] in , N(i) Indicates the relationship with the target node i The set of other connected source nodes, m i←j Indicates from the source node j To the target node i The message being delivered M i This represents the aggregation of messages passed from other source nodes to the target node.

[0098] After message aggregation is complete, the state representation of each message transmission channel is updated through the first gated loop unit, which can be represented as:

[0099]

[0100] in, M i This represents the aggregation of messages in node i. e ij This represents the current state of the message transmission channel between node i and node j. e’ ij This represents the updated state of the message transmission channel between node i and node j.

[0101] Finally, each node updates itself based on the state of its incoming edges. Node updates are performed using a second gated loop unit to ensure that historical information is not completely overwritten. Node updates can be represented as:

[0102]

[0103] in, h’ i This represents the updated node characteristics. Represents nodes The average of the updated state representations of the connected message transmission channels, w e→j This indicates that the node updates the weight matrix, where b is the bias term. h i This represents the characteristics of the current node.

[0104] After the node feature update is completed, the first loop ends. The message transmission goes through a preset number of iterations, allowing the nodes and edges of the initialized fully connected graph to be updated alternately, so as to realize the flow effect of the solar dark strip structure message and complete the global perception capability of each solar dark strip fragment node.

[0105] Finally, a multilayer perceptron is used to classify the edges between nodes of the solar filament fragments, thereby achieving a redistribution of the fragments' affiliations. The redistribution results are as follows: Figure 3 As shown.

[0106] Understandably, this embodiment utilizes the shape semantics, texture features, and spatial relationships of fragments, continuously exchanging information between nodes and edges through cyclic message passing. This gradually strengthens or suppresses potential connections between fragments, thereby achieving the classification and redistribution of dark fringe fragments. Fragments belonging to the same dark fringe are grouped together, resulting in an overall fringe morphology that more closely resembles the actual physical structure. Compared to traditional splicing methods that rely on geometric distance or manual rules, this scheme has stronger noise resistance, structural reasoning ability, and global perception capability. It can significantly improve the accuracy and robustness of dark fringe fragment association judgment, enabling more reliable solar activity structure identification and subsequent scientific research analysis.

[0107] It should be noted that the above embodiments are only for understanding this application and do not constitute a limitation on the solar filament fragment allocation method of this application. Any simple modifications based on this technical concept are within the protection scope of this application.

[0108] This application also provides a solar filament debris distribution device; please refer to... Figure 4 The solar filament debris distribution device includes:

[0109] Module 10 is used to acquire the full solar surface H-alpha image to be processed;

[0110] The feature extraction module 20 is used to input the H-alpha image of the full solar surface to be processed into the semantic segmentation model for solar dark stripe recognition and to obtain the binary mask data of the H-alpha image of the full solar surface to be processed.

[0111] The fully connected graph construction module 30 is used to determine the distribution of connected components of solar filaments based on binary mask data. It uses the connected component fragments of solar filaments as graph nodes and adds logical edges between any two nodes to construct an initial fully connected graph.

[0112] The node update module 40 is used to implement message passing between nodes in the initial connection graph, update node features and edge states, and obtain the updated fully connected graph based on the fragment node message passing mechanism.

[0113] The reallocation module 50 is used to classify the edges of the updated fully connected graph and, based on the classification results, reallocate the connected component fragments of the solar filament.

[0114] The solar filament fragment distribution device provided in this application, employing the solar filament fragment distribution method in the above embodiments, can solve the technical problem of insufficient accuracy in solar filament fragment distribution in related technologies. Compared with related technologies, the beneficial effects of the solar filament fragment distribution device provided in this application are the same as those of the solar filament fragment distribution method provided in the above embodiments, and other technical features in the solar filament fragment distribution device are the same as those disclosed in the methods of the above embodiments, and will not be repeated here.

[0115] This application provides a solar filament fragment allocation device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the solar filament fragment allocation method in the above embodiments.

[0116] The following is for reference. Figure 5 The diagram illustrates a structural schematic of a solar filament debris distribution device suitable for implementing embodiments of this application. The solar filament debris distribution device in this application may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Description), PMPs (Portable Media Players), and in-vehicle terminals (e.g., in-vehicle navigation terminals), as well as fixed terminals such as digital TVs and desktop computers. Figure 5 The solar filament fragment distribution device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.

[0117] like Figure 5As shown, the solar flare fragment allocation device may include a processing unit 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory 1002 (ROM) or a program loaded from a storage device 1003 into a random access memory 1004 (RAM). The random access memory 1004 also stores various programs and data required for the operation of the solar flare fragment allocation device. The processing unit 1001, the read-only memory 1002, and the random access memory 1004 are interconnected via a bus 1005. An input / output interface 1006 (I / O interface) is also connected to the bus 1005. Typically, the following systems can be connected to the input / output interface 1006: input devices 1007 including, for example, touchscreens, touchpads, keyboards, mice, image sensors, microphones, accelerometers, gyroscopes, etc.; output devices 1008 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 1003 including, for example, magnetic tapes, hard disks, etc.; and communication devices 1009. Communication device 1009 allows the solar flare debris distribution device to communicate wirelessly or wiredly with other devices to exchange data. Although solar flare debris distribution devices with various systems are shown in the figures, it should be understood that it is not required to implement or possess all of the systems shown. More or fewer systems may be implemented alternatively.

[0118] Although the diagrams show solar filament debris distribution devices with various systems, it should be understood that implementation or possession of all the systems shown is not required. More or fewer systems may be implemented alternatively.

[0119] Specifically, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from storage device 1003, or installed from ROM 1002. When the computer program is executed by processing device 1001, it performs the functions defined in the methods of the embodiments disclosed in this application.

[0120] The solar filament fragment allocation device provided in this application, employing the solar filament fragment allocation method described in the above embodiments, can solve the technical problem of insufficient accuracy in solar filament fragment allocation in related technologies. Compared with related technologies, the beneficial effects of the solar filament fragment allocation device provided in this application are the same as those of the solar filament fragment allocation method provided in the above embodiments, and other technical features in this solar filament fragment allocation device are the same as those disclosed in the previous embodiment method, and will not be repeated here.

[0121] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0122] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0123] This application provides a computer-readable storage medium having computer-readable program instructions (i.e., a computer program) stored thereon, which are used to execute the solar filament fragment allocation method in the above embodiments.

[0124] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, system, or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0125] The aforementioned computer-readable storage medium may be included in the solar filament fragment distribution device; or it may exist independently and not assembled into the solar filament fragment distribution device.

[0126] The aforementioned computer-readable storage medium carries one or more programs. When these programs are executed by the solar flare fragment allocation device, the solar flare fragment allocation device performs the following actions: acquires a full-plane H-alpha image to be processed; inputs the full-plane H-alpha image to be processed into a semantic segmentation model for solar flare identification, and obtains binary mask data of the full-plane H-alpha image to be processed; based on the binary mask data, determines the connected component distribution of the solar flares, and uses the connected component fragments of the solar flares as graph nodes, adds logical edges between any two nodes to construct an initial fully connected graph; based on the fragment node message passing mechanism, implements message passing between nodes in the initial connected graph, updates node features and edge states, and obtains an updated fully connected graph; classifies the edges of the updated fully connected graph, and based on the classification results, redistributes the connected component fragments of the solar flares.

[0127] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof. These programming languages ​​include object-oriented programming languages—such as Python, Java, Smalltalk, Python / C++—and conventional procedural programming languages—such as C or similar languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a Local Area Network (LAN) or a Wide Area Network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0128] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0129] The modules described in the embodiments of this application can be implemented in software or hardware. The names of the modules do not necessarily limit the functionality of the unit itself.

[0130] The readable storage medium provided in this application is a computer-readable storage medium that stores computer-readable program instructions (i.e., a computer program) for executing the above-described solar filament fragment allocation method, thereby solving the technical problem of insufficient accuracy in solar filament fragment allocation in related technologies. Compared with related technologies, the beneficial effects of the computer-readable storage medium provided in this application are the same as those of the solar filament fragment allocation method provided in the above embodiments, and will not be repeated here.

[0131] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the solar filament fragment allocation method described above.

[0132] The computer program product provided in this application can solve the technical problem of insufficient accuracy in the allocation of solar filament fragments in related technologies. Compared with related technologies, the beneficial effects of the computer program product provided in this application are the same as those of the solar filament fragment allocation method provided in the above embodiments, and will not be repeated here.

[0133] The above are only some embodiments of this application and do not limit the patent scope of this application. All equivalent structural transformations made under the technical concept of this application and using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included in the patent protection scope of this application.

Claims

1. A method for allocating solar filament fragments, characterized in that, The method includes: Obtain the full solar surface H-alpha image to be processed; The H-alpha image of the full solar surface to be processed is input into the semantic segmentation model to obtain the binary mask data of the H-alpha image of the full solar surface to be processed. Based on the binary mask data, the distribution of connected components of solar filaments is determined. Using the connected component fragments of solar filaments as graph nodes, logical edges are added between any two nodes to construct an initial fully connected graph. Based on the fragment node message passing mechanism, message passing between nodes in the initial connection graph is realized, node features and edge states are updated, and the updated fully connected graph is obtained. The edges of the updated fully connected graph are classified, and based on the classification results, the connected component fragments of the solar filament are redistributed. The steps of determining the connected component distribution of solar filaments based on the binary mask data, and constructing an initial fully connected graph by adding logical edges between any two nodes using the connected component fragments of solar filaments as graph nodes, include: For each connected component fragment in the binary mask data, a feature vector consistent with the location of the connected component fragment is determined in the high-level feature map output by the semantic segmentation model. The feature vectors of the connected component fragments are spatially mapped and aggregated to obtain high-dimensional representations of each connected component fragment. Based on a multilayer perceptron, the high-dimensional representation after nonlinear transformation is used as the node features of the initial fully connected graph to construct the initial fully connected graph; The steps for implementing message passing between nodes in the initial connected graph, updating node features and edge states, and obtaining the updated fully connected graph based on the fragment node message passing mechanism include: Centered on the edges of the initialized fully connected graph, for each edge, the state representation of each edge of the initialized fully connected graph is determined based on the node characteristics of the two endpoints of the edge; Using the edges of the initialized fully connected graph as message passing channels, the nonlinear transformation node features of the source node are concatenated with the state representation of the corresponding message passing channel to obtain the message to be passed from the source node. Through the corresponding message passing channel, message passing from the source node to the target node is realized, and message passing of all nodes in the initialized fully connected graph is completed. For each message passing channel, update the state representation of the message passing channel based on all messages flowing to the target node corresponding to the message passing channel; For each node in the initialized fully connected graph, the node in the initialized fully connected graph is updated based on the updated channel state of the message transmission channel connected to the node, completing one update cycle; Based on a preset number of iterations, the initial fully connected graph is updated cyclically to obtain the updated fully connected graph.

2. The method as described in claim 1, characterized in that, The step of determining the state representation of each edge of the initialized fully connected graph, centered on the edges, based on the node features after nonlinear transformation of the nodes at both ends of the edge, includes: For each edge, the node features after nonlinear transformation of the nodes at both ends of the edge are spliced ​​together to obtain the node splicing features. After linearly transforming the node splicing features based on the initial edge weight parameters, a nonlinear transformation is performed using the hyperbolic tangent activation function to obtain the initial state of the edge.

3. The method as described in claim 1, characterized in that, The step of updating the state representation of each message transmission channel based on all transmitted messages flowing to the target node corresponding to the message transmission channel includes: For each message passing channel, the state representation of the message passing channel is updated based on the first gated loop unit, using all the messages flowing to the target node corresponding to the message passing channel and the current state representation of the message passing channel as input.

4. The method as described in claim 3, characterized in that, The step of updating the nodes of the initialized fully connected graph for each node, based on the updated channel state of the message transmission channel connected to the node, includes: For each node of the initialized fully connected graph, the node of the initialized fully connected graph is updated based on the second gated loop unit, using the average of the updated channel states of all message transmission channels connected to the node and the node characteristics of the node of the initialized fully connected graph as input.

5. A solar streak fragment distribution device, characterized in that, The device includes: The acquisition module is used to acquire the full solar surface H-alpha image to be processed; The feature extraction module is used to input the H-alpha image of the full solar surface to be processed into the semantic segmentation model to obtain the binary mask data of the H-alpha image of the full solar surface to be processed. The fully connected graph construction module is used to determine the distribution of connected components of solar filaments based on the binary mask data, and to construct an initial fully connected graph by adding logical edges between any two nodes using the connected component fragments of solar filaments as graph nodes. The node update module is used to realize message passing between nodes in the initial connection graph based on the fragment node message passing mechanism, update node features and edge states, and obtain the updated fully connected graph. The reallocation module is used to classify the edges of the updated fully connected graph and, based on the classification results, reallocate the connected component fragments of the solar filament. The fully connected graph construction module is also used for: For each connected component fragment in the binary mask data, a feature vector consistent with the location of the connected component fragment is determined in the high-level feature map output by the semantic segmentation model. The feature vectors of the connected component fragments are spatially mapped and aggregated to obtain high-dimensional representations of each connected component fragment. Based on a multilayer perceptron, the high-dimensional representation after nonlinear transformation is used as the node features of the initial fully connected graph to construct the initial fully connected graph; The node update module is also used for: Centered on the edges of the initialized fully connected graph, for each edge, the state representation of each edge of the initialized fully connected graph is determined based on the node characteristics of the two endpoints of the edge; Using the edges of the initialized fully connected graph as message passing channels, the nonlinear transformation node features of the source node are concatenated with the state representation of the corresponding message passing channel to obtain the message to be passed from the source node. Through the corresponding message passing channel, message passing from the source node to the target node is realized, and message passing of all nodes in the initialized fully connected graph is completed. For each message passing channel, update the state representation of the message passing channel based on all messages flowing to the target node corresponding to the message passing channel; For each node in the initialized fully connected graph, the node in the initialized fully connected graph is updated based on the updated channel state of the message transmission channel connected to the node, completing one update cycle; Based on a preset number of iterations, the initial fully connected graph is updated cyclically to obtain the updated fully connected graph.

6. A solar filament fragment distribution device, characterized in that, The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the solar filament fragment allocation method as described in any one of claims 1 to 4.

7. A storage medium, characterized in that, The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, it implements the steps of the solar filament fragment allocation method as described in any one of claims 1 to 4.

8. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the steps of the solar filament fragment allocation method as described in any one of claims 1 to 4.

Images