Satellite-borne multi-computing board architecture-based ground surface emergency event detection method, device and medium

Abstract

The application provides a kind of ground surface emergency detection method, equipment and medium under spaceborne multi-computing board architecture, it is related to satellite remote sensing image processing technical field, comprising: remote sensing image data is sent to first computing module, second computing module and third computing module respectively;Make it to remote sensing image data carry out burst anomaly index extraction, obtain first burst anomaly index, second burst anomaly index and third burst anomaly index;Mixed expert computing module determines the weighting coefficient corresponding to first burst anomaly index, second burst anomaly index and third burst anomaly index according to the monitoring task corresponding to remote sensing image data, and at least one fusion strategy is used to weight and fuse first burst anomaly index, second burst anomaly index and third burst anomaly index, to obtain at least one ground surface emergency detection result.The technical scheme of the embodiment of the application utilizes multiple computing modules to process in parallel, improves the accuracy and detection efficiency of the detection result.

Description

TECHNICAL FIELD

[0001] The present application relates to the satellite remote sensing image processing technical field, especially to a satellite-borne multi-computing board architecture under ground surface emergency detection method, device and medium. BACKGROUND

[0002] At present, the accuracy requirement of remote sensing image processing is higher and higher, in the detection process of ground surface emergency (such as natural disasters, environmental pollution, etc.), the remote sensing image of different wave bands is analyzed respectively, so as to determine the result.

[0003] The traditional processing method is basically using single computing board platform for processing, in the case of rapid growth of current remote sensing data, the processing efficiency is low, in order to improve the processing efficiency, part of the dimensional information has to be abandoned, so that the detection result is not accurate enough. SUMMARY

[0004] Therefore, the purpose of the present application is to provide a satellite-borne multi-computing board architecture under ground surface emergency detection method, device and medium, through the parallel processing of multiple computing modules, the remote sensing image data is extracted from different dimensions respectively, the efficiency of ground surface emergency detection is improved, the extracted emergency abnormal index is weighted and fused by using different fusion strategies, the advantages of multiple computing boards are fully utilized, the high-precision detection of complex ground surface emergency is realized, and the accuracy of ground surface emergency detection result is improved.

[0005] In the first aspect, the present application embodiment provides a satellite-borne multi-computing board architecture under ground surface emergency detection method, which is applied to a satellite-borne multi-computing board architecture under ground surface emergency detection device, comprising: a remote sensing data source module, a first computing module, a second computing module, a third computing module and a hybrid expert computing module.

[0006] The method comprises:

[0007] The remote sensing data source module sends the remote sensing image data to the first computing module, the second computing module and the third computing module respectively;

[0008] The first computing module, the second computing module and the third computing module respectively extract the emergency abnormal index of the remote sensing image data, and obtain the first emergency abnormal index, the second emergency abnormal index and the third emergency abnormal index; wherein the first emergency abnormal index refers to the emergency abnormal index output by the first computing module, the second emergency abnormal index refers to the emergency abnormal index output by the second computing module, and the third emergency abnormal index refers to the emergency abnormal index output by the third computing module;

[0009] The mixed expert computing module determines the weighting coefficients corresponding to the first burst anomaly indicator, the second burst anomaly indicator and the third burst anomaly indicator respectively according to the monitoring task corresponding to the remote sensing image data, and performs weighted fusion on the first burst anomaly indicator, the second burst anomaly indicator and the third burst anomaly indicator according to at least one fusion strategy, to obtain at least one surface burst event detection result.

[0010] In a preferred embodiment of the present application, the number of remote sensing images is at least one frame; the first computing module, the second computing module and the third computing module respectively perform burst anomaly indicator extraction on each remote sensing image data to obtain the first burst anomaly indicator, the second burst anomaly indicator and the third burst anomaly indicator corresponding to each remote sensing image data, including:

[0011] For each frame of remote sensing image data, the first computing module performs burst anomaly indicator extraction on the remote sensing image data according to a remote anomaly index calculation model to obtain a first burst anomaly indicator, and sends interrupt information to the mixed expert computing module; the first burst anomaly indicator includes a normalized vegetation index, a normalized difference water index and a difference vegetation index;

[0012] For each frame of remote sensing image data, the second computing module performs burst anomaly indicator extraction on the remote sensing image data according to a spatial texture feature extraction model to obtain a second burst anomaly indicator, and sends interrupt information to the mixed expert computing module; the second burst anomaly indicator includes a co-occurrence matrix and a binary code;

[0013] For each frame of remote sensing image data, the third computing module performs burst anomaly indicator extraction on the remote sensing image data according to a deep feature extraction model to obtain a third burst anomaly indicator, and sends interrupt information to the mixed expert computing module; the third burst anomaly indicator includes color, shape and semantic information of a surface burst event.

[0014] In a preferred embodiment of the present application, the mixed expert computing module determines the weighting coefficients corresponding to the first burst anomaly indicator, the second burst anomaly indicator and the third burst anomaly indicator respectively according to the monitoring task corresponding to the remote sensing image data, including:

[0015] When the number of interrupt information received by the mixed expert computing module is greater than a preset number, the first burst anomaly indicator, the second burst anomaly indicator and the third burst anomaly indicator are read.

[0016] According to the monitoring task corresponding to the remote sensing image data, the weighting coefficients corresponding to the first burst anomaly indicator, the second burst anomaly indicator and the third burst anomaly indicator are queried.

[0017] In a preferred embodiment of the present application, the fusion strategy includes a weighted average strategy, a maximum voting strategy.

[0018] In a preferred embodiment of the present application, the mixed expert computing module includes an adaptive learning model.

[0019] The mixed expert computing module is configured to obtain the first burst anomaly indicator, the second burst anomaly indicator, the third burst anomaly indicator, a first historical burst anomaly indicator, a second historical burst anomaly indicator, and a third historical burst anomaly indicator; train the adaptive learning model according to the first burst anomaly indicator, the second burst anomaly indicator, the third burst anomaly indicator, the first historical burst anomaly indicator, the second historical burst anomaly indicator, and the third historical burst anomaly indicator, to obtain a trained adaptive learning model.

[0020] The trained adaptive learning model is configured to determine a weighting coefficient corresponding to the first burst anomaly indicator, the second burst anomaly indicator, and the third burst anomaly indicator according to a monitoring task corresponding to the remote sensing image data, and perform weighted fusion on the first burst anomaly indicator, the second burst anomaly indicator, and the third burst anomaly indicator according to at least one fusion strategy, to obtain at least one surface burst event detection result.

[0021] In a preferred embodiment of the present application, the first computing module, the second computing module, and the third computing module have the same structure, including an FPGA, a CPU, a GPU, and a memory; the FPGA, the CPU, and the GPU are connected in series, and the memory is connected to the FPGA and the CPU respectively.

[0022] The FPGA is configured to obtain each remote sensing image data, sequentially send the remote sensing image data to the CPU in a predetermined order, and send the first burst anomaly indicator, the second burst anomaly indicator, or the third burst anomaly indicator output by the CPU to the mixed expert computing module.

[0023] The CPU is configured to, for each frame of remote sensing image data, perform burst anomaly indicator extraction on the remote sensing image data, to obtain the first burst anomaly indicator, the second burst anomaly indicator, or the third burst anomaly indicator.

[0024] The GPU is configured to deploy the deep feature extraction model for the CPU to perform burst anomaly indicator extraction on the remote sensing image data, to obtain the third burst anomaly indicator.

[0025] The memory is configured to store, for each frame of remote sensing image data, the remote sensing image data, and the first burst anomaly indicator, the second burst anomaly indicator, or the third burst anomaly indicator.

[0026] In a preferred embodiment of the present application, the memory is further configured to clear the remote sensing image data and the first burst anomaly indicator, the second burst anomaly indicator, or the third burst anomaly indicator stored in the memory after the FPGA sends the first burst anomaly indicator, the second burst anomaly indicator, or the third burst anomaly indicator to the hybrid expert computing module.

[0027] In a preferred embodiment of the present application, the CPU further comprises an external network module configured to acquire the remote sensing image data through wireless communication.

[0028] In a second aspect, the embodiments of the present application further provide an electronic device comprising a processor and a memory, wherein the memory stores computer executable instructions capable of being executed by the processor, and the processor executes the computer executable instructions to implement the satellite-borne multi-computing board architecture-based ground surface burst event detection method of the first aspect.

[0029] In a fourth aspect, the embodiments of the present application further provide a computer readable storage medium storing computer executable instructions, wherein the computer executable instructions, when invoked and executed by a processor, cause the processor to implement the satellite-borne multi-computing board architecture-based ground surface burst event detection method of the first aspect.

[0030] The embodiments of the present application bring the following beneficial effects:

[0031] The embodiments of the present application provide a satellite-borne multi-computing board architecture-based ground surface burst event detection method, which is applied to a satellite-borne multi-computing board architecture-based ground surface burst event detection device. The first computing module, the second computing module, and the third computing module are used to simultaneously extract burst anomaly indicators from remote sensing image data, thereby obtaining a first burst anomaly indicator, a second burst anomaly indicator, and a third burst anomaly indicator. In this way, the analysis efficiency of the remote sensing image data is improved without reducing the analysis dimension. The hybrid expert computing module determines the weighting coefficients corresponding to the first burst anomaly indicator, the second burst anomaly indicator, and the third burst anomaly indicator according to different monitoring tasks, and fuses the first burst anomaly indicator, the second burst anomaly indicator, and the third burst anomaly indicator based on different fusion strategies. In this way, the advantages of different computing modules can be fully utilized for different detection tasks, high-precision detection of complex ground surface burst events is achieved, and the accuracy of ground surface burst event detection is improved.

[0032] Other features and advantages of the present application will be set forth in the descriptions that follow, and in part will be apparent from the description, or can be learned by practice of the application as hereinafter described, or can be learned by practice of the application. It is also to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the application, as claimed.

[0033] In order to make the above objectives, features and advantages of the present application more obvious and comprehensible, the following preferred embodiments are specifically described in detail below, and the accompanying drawings are referred to. BRIEF DESCRIPTION OF DRAWINGS

[0034] In order to more clearly illustrate the specific embodiments of the present application or the technical solutions in the prior art, the following will briefly introduce the drawings needed to be used in the specific embodiments or the prior art description. Obviously, the drawings described below are some embodiments of the present application, and those skilled in the art can obtain other drawings according to these drawings without any creative labor.

[0035] Figure 1a A structural schematic diagram of a ground surface emergency detection device under a spaceborne multi-computing board architecture provided by an embodiment of the present application;

[0036] Figure 1b A flowchart of a ground surface emergency detection method under a spaceborne multi-computing board architecture provided by an embodiment of the present application;

[0037] Figure 2 A flowchart of another ground surface emergency detection method under a spaceborne multi-computing board architecture provided by an embodiment of the present application;

[0038] Figure 3a A flowchart of another ground surface emergency detection method under a spaceborne multi-computing board architecture provided by an embodiment of the present application;

[0039] Figure 3b A structural diagram of a computing module provided by an embodiment of the present application;

[0040] Figure 4 A structural schematic diagram of an electronic device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0041] In order to make the objectives, technical solutions and advantages of the embodiments of the present application more clear, the technical solutions of the present application will be described clearly and completely below with reference to the drawings. Obviously, the described embodiments are some embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without any creative labor are within the protection scope of the present application.

[0042] The accuracy requirement of satellite remote sensing image processing is higher and higher, in the detection process of the ground surface emergency (such as natural disasters, environmental pollution, etc.), the feature extraction and analysis of remote sensing images of different bands are carried out respectively, and finally the fusion judgment result is obtained. However, the traditional processing method is basically processed by using a single computing board platform, and in the case of rapid growth of current remote sensing data, the processing efficiency is low, and in order to improve the processing efficiency, part of the dimensional information has to be abandoned, thereby causing the detection result to be not accurate enough.

[0043] Based on this, the ground surface emergency detection method under the satellite multi-computing board architecture provided by the embodiment of the application can extract the information of multiple dimensions of remote sensing image data through multiple computing modules at the same time, so as to achieve the purpose of generating accurate and robust ground surface anomaly detection results more efficiently without abandoning different dimensional information.

[0044] In order to facilitate the understanding of the embodiment, first, the ground surface emergency detection method under the satellite multi-computing board architecture disclosed by the embodiment of the application is introduced in detail.

[0045] Embodiment 1

[0046] The ground surface emergency detection method under the satellite multi-computing board architecture provided by the embodiment of the application is applied to a ground surface emergency detection device under the satellite multi-computing board architecture, which comprises a remote sensing data source module, a first computing module, a second computing module, a third computing module and a hybrid expert computing module. Figure 1a The structure diagram of the ground surface emergency detection device under the satellite multi-computing board architecture provided by the embodiment of the application is shown in FIG. 1. Figure 1a As shown in the figure, the device comprises a hybrid expert computing unit, a computing board 1, a computing board 2, a computing board 3, a remote sensing data source and an interface backboard. The remote sensing data source module runs in the remote sensing data source, the first computing module runs in the computing board 1, the second computing module runs in the computing board 2, the third computing module runs in the computing board 3, and the hybrid expert computing module runs in the hybrid expert computing unit. The hybrid expert computing unit, the computing board 1, the computing board 2, the computing board 3 and the remote sensing data source are respectively inserted into different slots of the same interface backboard, and the device can be installed on a remote sensing satellite through the interface backboard to form a satellite multi-computing board architecture. The positions of the hybrid expert computing unit, the computing board 1, the computing board 2, the computing board 3 and the remote sensing data source can be exchanged with each other. Figure 1b The flow chart of the ground surface emergency detection method under the satellite multi-computing board architecture provided by the embodiment of the application is shown in FIG. 2. Figure 1b As shown in the figure, the ground surface emergency detection method under the satellite multi-computing board architecture can comprise the following steps:

[0047] Step S101, the remote sensing data source module sends remote sensing image data to the first computing module, the second computing module and the third computing module respectively.

[0048] The remote sensing data source module refers to a device for storing remote sensing image data. In the remote sensing data source module, a plurality of frames of remote sensing image data are stored. The remote sensing data source can send the remote sensing image data to the first computing module, the second computing module and the third computing module in a certain order. The remote sensing data source module can send the next frame of remote sensing image data after the first computing module, the second computing module and the third computing module complete processing of a frame of remote sensing image data, or can send a plurality of frames of remote sensing image data to the first computing module, the second computing module and the third computing module in a certain order. The first computing module, the second computing module and the third computing module process in turn according to the order of the plurality of frames of remote sensing image data. When the remote sensing data source module sends the plurality of frames of remote sensing image data to the first computing module, the second computing module and the third computing module, the number of frames of remote sensing image data sent is determined by the memory size of the first computing module, the second computing module and the third computing module, that is, if the memory of each of the three computing modules does not overflow during processing, the next frame of remote sensing image data to be processed can be pre-read from the remote sensing data source module. This processing method avoids waiting delay and improves the real-time performance of the entire processing flow.

[0049] Step S102, the first computing module, the second computing module and the third computing module respectively extract the burst anomaly index of the remote sensing image data to obtain the first burst anomaly index, the second burst anomaly index and the third burst anomaly index.

[0050] The first computing module, the second computing module and the third computing module simultaneously perform burst anomaly index extraction on the remote sensing image data, and respectively obtain a first burst anomaly index, a second burst anomaly index and a third burst anomaly index. The first burst anomaly index refers to the burst anomaly index output by the first computing module, the second burst anomaly index refers to the burst anomaly index output by the second computing module, and the third burst anomaly index refers to the burst anomaly index output by the third computing module. The first burst anomaly index refers to a remote sensing index capable of reflecting abnormal information of the ground in the remote sensing image data. The second burst anomaly index is used to describe the structural information of the ground. The third burst anomaly index is used to describe the characteristics from low to high in the ground burst event, including color, shape and semantic information of the abnormal event, etc. Through the first burst anomaly index, the second burst anomaly index and the third burst anomaly index, the abnormal situation in the remote sensing image data can be reflected from different aspects. The ground burst event refers to various phenomena or events occurring on the earth's surface, which are different from normal natural phenomena or human activity patterns. Examples include forest fires, water pollution, land desertification, urban expansion, mining, etc.

[0051] In step S103, the mixed expert computing module determines the weighting coefficients corresponding to the first burst anomaly index, the second burst anomaly index and the third burst anomaly index according to the monitoring task corresponding to the remote sensing image data, and performs weighted fusion on the first burst anomaly index, the second burst anomaly index and the third burst anomaly index according to at least one fusion strategy to obtain at least one ground burst event detection result.

[0052] The monitoring task is used to describe the specific target or phenomenon that needs to be detected and analyzed through the remote sensing image data, which determines the direction and focus of the entire data processing and analysis. The monitoring task can be pre-set or matched by looking up the table according to the ground area collected by the remote sensing image data. Different monitoring tasks have different attention degrees to different burst anomaly indexes, so the weighting coefficients are different. For example, when the monitoring task is forest fire monitoring, the attention degree to the first burst anomaly index is the highest, the attention degree to the third burst anomaly index is the second, and the attention degree to the second burst anomaly index is the smallest. At this time, the weighting coefficient of the first burst anomaly index is the largest, the weighting coefficient of the third burst anomaly index is the second, and the weighting coefficient of the second burst anomaly index is the smallest.

[0053] The fusion strategy refers to how to fuse the weighted first burst anomaly index, the second burst anomaly index and the third burst anomaly index. Different fusion strategies obtain different ground burst event detection results.

[0054] Specifically, the mixed expert computing module can query the first burst anomaly indicator, the second burst anomaly indicator and the third burst anomaly indicator corresponding to the weighting coefficients respectively according to the monitoring task corresponding to the remote sensing image data in a table lookup manner. For the weighted first burst anomaly indicator, the weighted second burst anomaly indicator and the weighted third burst anomaly indicator, at least one fusion strategy is used for fusion, and each fusion strategy obtains a ground surface burst event detection result.

[0055] The fusion strategy can include weighted average, maximum voting method or model-based integration method. The weighted average refers to summing the weighted first burst anomaly indicator, the weighted second burst anomaly indicator and the weighted third burst anomaly indicator, and then averaging, and according to the calculated average value, the average value corresponding to the ground surface burst event detection result is queried. The maximum voting method refers to voting the weighted first burst anomaly indicator, the weighted second burst anomaly indicator and the weighted third burst anomaly indicator multiple times, and finally outputting the most possible ground surface burst event. That is, the mixed expert computing module respectively determines the possible ground surface burst event for the weighted first burst anomaly indicator, the weighted second burst anomaly indicator and the weighted third burst anomaly indicator, and votes, and after a preset number of votes, the voting results are counted to find the ground surface burst event with the most votes as the ground surface burst event detection result.

[0056] The embodiment of the application provides a ground surface burst event detection method under a spaceborne multi-computing board architecture. The method is applied to a ground surface burst event detection device under a spaceborne multi-computing board architecture. The first computing module, the second computing module and the third computing module simultaneously extract burst anomaly indicators from remote sensing image data to obtain the first burst anomaly indicator, the second burst anomaly indicator and the third burst anomaly indicator respectively. In the case of not reducing the analysis dimension, the analysis efficiency of the remote sensing image data is improved. The mixed expert computing module determines the weighting coefficients corresponding to the first burst anomaly indicator, the second burst anomaly indicator and the third burst anomaly indicator respectively according to different monitoring tasks, and fuses the first burst anomaly indicator, the second burst anomaly indicator and the third burst anomaly indicator based on different fusion strategies. The advantages of different computing modules can be fully played for different detection tasks, high-precision detection of complex ground surface burst events is realized, and the accuracy of ground surface burst event detection is improved.

[0057] Embodiment 2

[0058] The embodiment of the present application also provides another surface burst event detection method under a spaceborne multi-computing board architecture; the method is realized on the basis of the method in the above embodiment; the method mainly describes the specific implementation mode of the first computing module, the second computing module and the third computing module for extracting the burst anomaly index of each remote sensing image data, to obtain the first burst anomaly index, the second burst anomaly index and the third burst anomaly index corresponding to each remote sensing image data.

[0059] Figure 2 The flow chart of another surface burst event detection method under a spaceborne multi-computing board architecture provided by the embodiment of the present application is shown in Figure 2 The surface burst event detection method under a spaceborne multi-computing board architecture can include the following steps:

[0060] In step S201, the remote sensing data source module sends remote sensing image data to the first computing module, the second computing module and the third computing module respectively.

[0061] In step S202, for each frame of remote sensing image data, the first computing module extracts the burst anomaly index according to the remote sensing anomaly index calculation model, obtains the first burst anomaly index, and sends the interrupt information to the hybrid expert computing module.

[0062] The remote sensing anomaly index model focuses on extracting the remote sensing index that can reflect the surface anomaly information in the remote sensing image. It can be understood that different wave bands provide diversified information of the ground. For example, the RGB wave band helps to capture the changes of the ground color, texture and geometric shape, and the infrared wave band can reveal the temperature and vegetation health characteristics. The remote sensing anomaly index calculation model mainly calculates the ratio or difference of different wave band images, so as to obtain the numerical index reflecting the ground features, that is, the first burst anomaly index.

[0063] The first burst anomaly index includes at least one of the normalized vegetation index, the normalized difference water body index and the difference vegetation index. The normalized difference vegetation index (NDVI) is used to evaluate the density and health degree of vegetation, and the NDVI can be calculated by the following formula:

[0064]

[0065] Wherein, NIR represents the near-infrared wave band, and Red represents the red light wave band. The range of NDVI value is [-1, 1], and the higher the NDVI value, the more lush the vegetation.

[0066] Normalized Difference Water Index (NDWI) is used to detect the presence of water bodies in the surface, especially in urbanized and arid regions. NDWI can be calculated by the following formula:

[0067]

[0068] where Green represents the green band, and the NDWI value ranges from [-1, 1], and the NDWI value of water body area is usually high.

[0069] Difference Vegetation Index (DVI) is used to evaluate the growth status of vegetation. DVI can be calculated by the following formula:

[0070] DVI = NIR - Red

[0071] where DVI is commonly used to distinguish between vegetation and non-vegetation areas, especially in low vegetation coverage areas.

[0072] Further, the calculation of the normalized vegetation index, the normalized difference water index, and the difference vegetation index can be processed in series or in parallel in the first calculation module to improve processing efficiency.

[0073] Specifically, for each frame of remote sensing image data, the first calculation module inputs the remote sensing image data into the remote sensing anomaly index calculation model to extract the burst anomaly index of the remote sensing image data, obtains the first burst anomaly index output by the remote sensing anomaly index calculation model, and sends an interrupt information to the hybrid expert calculation module. The interrupt information is used to inform the hybrid expert calculation module that the processing of one frame of remote sensing image data has been completed.

[0074] Step S203, for each frame of remote sensing image data, the second calculation module extracts the burst anomaly index of the remote sensing image data according to the spatial texture feature extraction model, obtains the second burst anomaly index, and sends an interrupt information to the hybrid expert calculation module.

[0075] The spatial texture feature extraction model captures the structural information of the surface by analyzing the spatial texture features of the features in the remote sensing image. Spatial texture features are crucial for identifying surface anomalies (such as urban expansion, farmland changes, etc.). Spatial texture features can be described by image gray level co-occurrence matrix (Gray-Level Co-occurrence Matrix, GLCM) and local binary patterns (Local Binary Patterns, LBP).

[0076] GLCM captures the texture information of ground objects by calculating the spatial relationship between pixel gray values. It counts the frequency of pixel pairs with specific gray values appearing in a given direction and distance, which is used to describe the co-occurrence relationship between pixel gray value pairs in the image, and then extracts features such as contrast, correlation, and uniformity. GLCM features include contrast (reflecting the degree of gray change, higher contrast indicating greater gray difference in the image), correlation (reflecting the degree of linear correlation between pixel pairs, high correlation indicating more consistent texture in the image), and uniformity (indicating the uniformity of gray distribution in the image, higher uniformity indicating smoother texture).

[0077] The contrast CON can be calculated by the following formula:

[0078]

[0079] where p(i,j) is the probability of pixel pairs with gray values i and j appearing in the gray level co-occurrence matrix, and N is the number of gray levels in the image.

[0080] The correlation COR can be calculated by the following formula:

[0081]

[0082] where μ i is the mean of gray value i, μ j is the mean of gray value j, σ i is the standard deviation of gray value i, and σ j is the standard deviation of gray value j.

[0083] The uniformity ASM can be calculated by the following formula:

[0084]

[0085] LBP generates a texture feature map by comparing the gray value of each pixel with the gray value of the surrounding neighborhood pixels to form a binary pattern. LBP can effectively capture the local texture features of the image and is suitable for images with obvious texture patterns.

[0086] where, for a center pixel with gray value g c , the surrounding neighborhood pixel gray values are g1, g2, g3, …, g8, and LBP can be calculated by the following formula:

[0087]

[0088] where s(x) is the sign function, s(x) = 1 when x ≥ 0, and s(x) = 0 when x < 0.

[0089] Specifically, for each frame of remote sensing image data, the second computing module inputs the remote sensing image data into the spatial texture feature extraction model, extracts the burst anomaly index of the remote sensing image data, obtains the second burst anomaly index output by the spatial texture feature extraction model, and sends the interrupt information to the hybrid expert computing module.

[0090] Further, the calculation of the image gray level co-occurrence matrix and the local binary pattern can be processed in series or in parallel in the second computing module to improve the processing efficiency.

[0091] Step S204, for each frame of remote sensing image data, the third computing module extracts the burst anomaly index of the remote sensing image data according to the deep feature extraction model, obtains the third burst anomaly index, and sends the interrupt information to the hybrid expert computing module.

[0092] The deep feature extraction model uses a deep learning model, such as a convolutional neural network (CNN), a fully convolutional neural network (FCN), and a graph convolutional neural network (GCN), to extract multi-level features from remote sensing images. The deep learning model can automatically extract low-level to high-level features from surface burst events, including color, shape, and semantic information of the abnormal event, i.e., the third burst anomaly index. In addition, in order to reduce the computational consumption and complexity of the model, the model can be lightweighted by using model pruning, knowledge distillation, etc. strategy, so that the deep feature extraction model provides efficient and high-precision surface anomaly detection.

[0093] Specifically, for each frame of remote sensing image data, the third computing module inputs the remote sensing image data into the deep feature extraction model, extracts the burst anomaly index of the remote sensing image data, obtains the third burst anomaly index output by the deep feature extraction model, and sends the interrupt information to the hybrid expert computing module.

[0094] Step S205, the hybrid expert computing module determines the weighting coefficients corresponding to the first burst anomaly index, the second burst anomaly index and the third burst anomaly index according to the monitoring task corresponding to the remote sensing image data, and performs weighted fusion on the first burst anomaly index, the second burst anomaly index and the third burst anomaly index according to at least one fusion strategy, to obtain at least one surface burst event detection result.

[0095] Further, the mixed expert computing module determines the weighting coefficients corresponding to the first burst anomaly indicator, the second burst anomaly indicator and the third burst anomaly indicator according to the monitoring task corresponding to the remote sensing image data, including: when the number of interruption information received by the mixed expert computing module is greater than a preset number, reading the first burst anomaly indicator, the second burst anomaly indicator and the third burst anomaly indicator; and querying the weighting coefficients corresponding to the first burst anomaly indicator, the second burst anomaly indicator and the third burst anomaly indicator according to the monitoring task corresponding to the remote sensing image data.

[0096] wherein the preset number is used to determine whether the first computing model, the second computing model and the third computing model have completed burst anomaly indicator extraction on the same frame of remote sensing image data; when the number of interruption information is greater than the preset number, it indicates that the first computing model, the second computing model and the third computing model have completed burst anomaly indicator extraction on the same frame of remote sensing image data; at this time, the mixed expert computing module reads the first burst anomaly indicator, the second burst anomaly indicator and the third burst anomaly indicator output by the first computing model, the second computing model and the third computing model respectively. According to the monitoring task corresponding to the remote sensing image data, the weighting coefficients corresponding to the first burst anomaly indicator, the second burst anomaly indicator and the third burst anomaly indicator under the detection task are queried by table lookup. In the embodiment of the application, the preset number is 3.

[0097] By setting the interruption information and counting the number of interruption information, it can be ensured that the first burst anomaly indicator, the second burst anomaly indicator and the third burst anomaly indicator obtained by the mixed expert computing module are burst anomaly indicators of the same frame of remote sensing image data.

[0098] Further, the mixed expert computing module comprises an adaptive learning model; the mixed expert computing module is configured to obtain the first burst anomaly indicator, the second burst anomaly indicator, the third burst anomaly indicator, a first historical burst anomaly indicator, a second historical burst anomaly indicator and a third historical burst anomaly indicator; train the adaptive learning model according to the first burst anomaly indicator, the second burst anomaly indicator, the third burst anomaly indicator, the first historical burst anomaly indicator, the second historical burst anomaly indicator and the third historical burst anomaly indicator, to obtain a trained adaptive learning model; and the trained adaptive learning model is configured to determine the weighting coefficients corresponding to the first burst anomaly indicator, the second burst anomaly indicator and the third burst anomaly indicator according to the monitoring task corresponding to the remote sensing image data, and perform weighted fusion on the first burst anomaly indicator, the second burst anomaly indicator and the third burst anomaly indicator according to at least one fusion strategy, to obtain at least one surface burst event detection result.

[0099] The first historical burst anomaly indicator refers to the first burst anomaly indicator output by the first calculation module in a historical time period. The second historical burst anomaly indicator refers to the second burst anomaly indicator output by the second calculation module in the historical time period. The third historical burst anomaly indicator refers to the third burst anomaly indicator output by the third calculation module in the historical time period. It can be understood that the adaptive learning model can be trained according to the first anomaly data, the second anomaly data and the third anomaly data output by the first calculation module, the second calculation module and the third calculation module respectively at the current moment, so that the adaptive learning model can be updated and optimized in real time to ensure that the optimal detection result is always provided under different types of remote sensing image data or different scenes.

[0100] The method for detecting surface burst events under the star-borne multi-computing board architecture provided by the embodiment of the application can analyze remote sensing anomaly indexes, spatial texture features, color, shape and semantic information of abnormal events and other indexes through the first calculation module, the second calculation module and the third calculation module, can determine surface burst events from multiple dimensions, and improves the accuracy of the detection result of surface burst events.

[0101] Embodiment 3

[0102] The embodiment of the application further provides another method for detecting surface burst events under the star-borne multi-computing board architecture; the method is implemented on the basis of the method in the above embodiment.

[0103] Figure 3a The flow chart of another method for detecting surface burst events under the star-borne multi-computing board architecture provided by the embodiment of the application is shown in FIG. 3, which can include the following steps: Figure 3a

[0104] Step S301, initialize each task module.

[0105] ​Specifically, the hybrid expert computing module is a separate module on hardware, and also serves as a master control module of task scheduling to control the entire ground surface emergency detection method process under the on-board multi-computing board architecture. First, the functional modules of each single board performing task processing are initialized, that is, the functional modules of the first computing module, the second computing module and the third computing module are initialized. Since image data needs to be cached at some specific steps in the image processing process, it is necessary to ensure that the required capacity of the related memory is released to prepare for storing the data in the processing process. Second, the master control module monitors the entire ground surface emergency detection method process under the on-board multi-computing board architecture, especially needs special processing when the data flow is synchronized. The core mechanism is to control the data flow to prevent memory overflow or data coverage caused by insufficient memory capacity, resulting in unavailable data. Finally, in order to achieve higher data transmission efficiency, the high-speed serial transceiver resources (GTH / GTX) of FPGA are used for data interaction between single boards, the interface protocol can be flexibly configured (such as Aurora, SRIO, etc.), and the data interface between the CPU and the FPGA in each single board uses PCIe, wherein the H2C and C2H channels are used as data download and upload channels, and the AXI-Lite channel is used as an instruction issuing and parameter configuration channel.

[0106] It can be understood that the first computing module, the second computing module and the third computing module have the same structure, which includes: FPGA, CPU, GPU, memory; the FPGA, the CPU and the GPU are connected in series, and the memory is connected with the FPGA and the CPU respectively. Figure 3b A structural diagram of a computing module provided by the embodiment of the present application is shown in Figure 3bAs shown, the FPGA is configured to acquire each of the remote sensing image data, sequentially send the remote sensing image to the CPU according to a predetermined order, and send the first burst anomaly index, the second burst anomaly index or the third burst anomaly index output by the CPU to the hybrid expert computing module; wherein TX and RX are transmission interfaces of the FPGA for acquiring remote sensing images. The CPU is configured to extract a burst anomaly index from the remote sensing image data for each frame of remote sensing image data, to obtain the first burst anomaly index, the second burst anomaly index or the third burst anomaly index; the GPU is configured to deploy the deep feature extraction model for the CPU to extract a burst anomaly index from the remote sensing image data, to obtain the third burst anomaly index; the memory is configured to store the remote sensing image data and the first burst anomaly index, the second burst anomaly index or the third burst anomaly index for each frame of remote sensing image data. The CPU further comprises an external network module configured to acquire the remote sensing image data through wireless communication. The memory comprises at least one high-efficiency DDR memory (Double Data Rate, double data rate synchronous dynamic random access memory) and a hard disk, wherein the DDR is configured to store data generated during the running of the FPGA or the CPU. Since the storage space of the DDR is small, the hard disk can be used to store the data generated during the running of the CPU.

[0107] Step S302, determine whether each functional module is ready, if yes, execute step S303, if no, execute step S301.

[0108] Specifically, the hybrid expert computing module determines whether the first computing module, the second computing module and the third computing module are ready.

[0109] Step S303, determine whether each computing board cache is ready, if yes, execute step S304, if no, execute step S303.

[0110] Specifically, each computing board, i.e. the first computing module, the second computing module and the third computing module, determines whether the memory is ready.

[0111] Step S304, write the remote sensing image data into each computing unit cache.

[0112] Specifically, the computing unit cache refers to the memory in the computing board. The remote sensing data is synchronously read into the first computing module, the second computing module and the third computing module, and the same frame of remote sensing image data is processed by the three computing boards.

[0113] Step S305, image slicing.

[0114] Specifically, the remote sensing image data has four bands, R, G, B and Nir bands. The size of a single frame of remote sensing image data can be customized as needed. For example, the width of the remote sensing image data is defined as W, and the height is defined as H. When there is a need for cutting, the slice width is defined as w, and the slice height is defined as h. Wherein, W is an integer multiple of w, and H is an integer multiple of h.

[0115] In each computing board, the remote sensing image data can be sliced first to obtain a plurality of slice images.

[0116] Step S306, simultaneously solve the normalized vegetation index, the normalized difference water index and the difference vegetation index.

[0117] Specifically, in the first computing module, after the FPGA obtains the plurality of frames of remote sensing image data, the remote sensing image data can be stored in the memory in priority, and the remote sensing image data is sent to the CPU in the preset order frame by frame. The CPU extracts the burst anomaly index of the remote sensing image data to obtain the first burst anomaly index, that is, the normalized vegetation index, the normalized difference water index and the difference vegetation index.

[0118] Step S307, the first burst anomaly index is cached and the master module is interrupted.

[0119] Specifically, the first computing module caches the first burst anomaly index in the memory and sends the interrupt information to the master module.

[0120] Step S308, simultaneously solve the co-occurrence matrix and binary coding.

[0121] Specifically, in the second computing module, after the FPGA obtains the plurality of frames of remote sensing image data, the remote sensing image data can be stored in the memory in priority, and the remote sensing image data is sent to the CPU in the preset order frame by frame. The CPU extracts the burst anomaly index of the remote sensing image data to obtain the second burst anomaly index, that is, the co-occurrence matrix and the binary coding.

[0122] Step S309, the second burst anomaly index is cached and the master module is interrupted.

[0123] Specifically, the second computing module caches the second burst anomaly index in the memory and sends the interrupt information to the master module.

[0124] Step S310, neural network inference deep features.

[0125] Specifically, in the third computing module, after the FPGA obtains the multiple frames of remote sensing image data, the remote sensing image data can be preferentially stored in the memory, and the remote sensing image data is sent to the CPU frame by frame in a preset order, the CPU extracts the burst anomaly index by using the deep learning model deployed in the GPU to obtain the third burst anomaly index, that is, the deep feature, wherein the deep feature includes color, shape and semantic information of the abnormal event.

[0126] In step S311, the third burst anomaly index is cached and an interrupt is notified to the master module.

[0127] Specifically, the third computing module caches the third burst anomaly index in the memory and sends the interrupt information to the master module.

[0128] In step S312, the interrupt is received and counted.

[0129] Specifically, the master module counts the number of received interrupt information.

[0130] In step S313, it is determined whether the number of interrupts is greater than N, if yes, step S314 is executed, and if no, step S312 is executed.

[0131] Specifically, the master module determines whether the number of interrupts is greater than N. N is a multiple of 3, and in the initial state, the multiple is 1, that is, N is 3. When the number of interrupt information is greater than N, the multiple can be increased by 1, and N can be re-determined, that is, N is re-determined as 6, and so on.

[0132] In step S314, the calculation result data of each computing board is read.

[0133] Specifically, the master module reads the first abnormal index, the second abnormal index and the third abnormal index from the first computing module, the second computing module and the third computing module respectively.

[0134] In step S315, it is determined whether the single frame data processing result is read. If yes, steps S316 and S317 are executed respectively, and if no, step S314 is executed.

[0135] Specifically, the master module determines whether the first abnormal index, the second abnormal index and the third abnormal index are read.

[0136] In step S316, the adaptive feature weighted fusion is performed.

[0137] Specifically, the master module determines the weighting coefficients corresponding to the first burst anomaly index, the second burst anomaly index and the third burst anomaly index according to the monitoring task corresponding to the remote sensing image data, and performs weighted fusion on the first burst anomaly index, the second burst anomaly index and the third burst anomaly index according to at least one fusion strategy.

[0138] Step S317, notify each computing board to release the cache occupied by the processed data.

[0139] Specifically, the host module notifies each computing board to release the cache occupied by the processed data after reading the first, second and third anomaly indexes. It can be understood that in each computing board, the memory is also used to clear the remote sensing image data and the first, second or third anomaly index stored in the memory after the FPGA sends the first, second or third anomaly index to the hybrid expert computing module.

[0140] Step S318, output the surface anomaly result.

[0141] Specifically, the host module outputs the surface anomaly result, which includes at least one surface burst event detection result.

[0142] Step S319, determine whether a task end instruction is received, if yes, execute step S320, and if no, execute step S304.

[0143] Specifically, if all remote sensing image data in the remote sensing data source module has been cached in each computing board, the host module will receive a task end instruction, otherwise, it will continue to obtain remote sensing image data from the remote sensing data source module.

[0144] Step S320, reset each functional module after outputting the result data of the last frame of remote sensing image.

[0145] Specifically, each functional module is reset after outputting the surface burst event detection result of the last frame of remote sensing image.

[0146] The embodiment of the application provides a surface burst event detection method under a spaceborne multi-computing board architecture. The method is applied to a surface burst event detection device under the spaceborne multi-computing board architecture. The first, second and third anomaly indexes are obtained by simultaneously extracting the anomaly indexes from the remote sensing image data through the first, second and third computing modules. The analysis efficiency of the remote sensing image data is improved without reducing the analysis dimension. The hybrid expert computing module determines the weighting coefficients corresponding to the first, second and third anomaly indexes according to different monitoring tasks, and fuses the first, second and third anomaly indexes based on different fusion strategies. The advantages of different computing modules can be fully utilized for different detection tasks to realize high-precision detection of complex surface burst events and improve the accuracy of surface burst event detection.

[0147] Embodiment 4

[0148] The embodiment of the present application also provides an electronic device for running the ground burst detection method under the spaceborne multi-computing board architecture. Figure 4 As shown in a structural schematic diagram of an electronic device, the electronic device comprises a memory 400 and a processor 401, wherein the memory 400 is used for storing one or more computer instructions, and the one or more computer instructions are executed by the processor 401 to realize the ground burst detection method under the spaceborne multi-computing board architecture.

[0149] Further, Figure 4 As shown in the electronic device, the electronic device further comprises a bus 402 and a communication interface 403, and the processor 401, the communication interface 403 and the memory 400 are connected through the bus 402.

[0150] The memory 400 can contain a high-speed random access memory (RAM) and can also include a non-volatile memory, for example, at least one disk memory. The communication connection between the system network element and at least one other network element is realized through at least one communication interface 403 (which can be wired or wireless), and the Internet, a wide area network, a local area network, a metropolitan area network, etc. can be used. The bus 402 can be an ISA bus, a PCI bus or an EISA bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For the convenience of representation, Figure 4 In the figure, only one bidirectional arrow is used to represent the bus, but it does not mean that there is only one bus or only one type of bus.

[0151] The processor 401 can be an integrated circuit chip having a signal processing capability. In the implementation process, each step of the above method can be completed by the integrated logic circuit of hardware in the processor 401 or the instruction in the form of software. The processor 401 described above can be a general processor, including a central processing unit (CPU), a network processor (NP), etc.; can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic device, a discrete gate or transistor logic device, a discrete hardware component. Each method, step and logic block diagram disclosed in the embodiment of the present application can be implemented or executed. The general processor can be a microprocessor or the processor can also be any conventional processor. The steps of the method disclosed in combination with the embodiment of the present application can be directly embodied as a hardware code processor for execution, or a combination of hardware and software modules in the code processor for execution. The software module can be located in a random access memory, a flash memory, a read-only memory, a programmable read-only memory or an electrically erasable programmable memory, a register, etc. The storage medium in the art. The storage medium is located in the memory 400, and the processor 401 reads the information in the memory 400, and combines the hardware to complete the steps of the method of the above embodiment.

[0152] The embodiment of the present application also provides a computer readable storage medium, the computer readable storage medium stores computer executable instructions, when the computer executable instructions are called and executed by the processor, the computer executable instructions cause the processor to implement the above-mentioned ground burst event detection method under the spaceborne multi-computing board architecture. For specific implementation, please refer to the method embodiment, which will not be repeated here.

[0153] The computer program product for performing the ground burst event detection method under the spaceborne multi-computing board architecture provided by the embodiment of the present application includes a computer readable storage medium storing non-volatile program codes executable by the processor. The instructions included in the program codes can be used to execute the method described in the foregoing method embodiment. For specific implementation, please refer to the method embodiment, which will not be repeated here.

[0154] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the system, device and unit described above can refer to the corresponding process in the foregoing method embodiment, which will not be repeated here.

[0155] In several embodiments provided by the present application, it should be understood that the disclosed system, device and method can be implemented in other manners. The described device embodiments are merely schematic, and for example, the division of the units is only a logical function division, and there can be another division manner in actual implementation; for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the displayed or discussed mutual couplings or direct couplings or communication connections between different units, or the among different units, can be indirect couplings or communication connections through some interfaces, devices or units, and can be in electric, mechanical or other forms.

[0156] The units described as separated components can or can not be physically separated, and the components displayed as units can or can not be physical units, i.e., can be located in one place, or can be distributed on a plurality of network units. In actual implementation, some or all of the units can be selected according to the actual needs to achieve the purposes of the embodiments of the present application.

[0157] In addition, each function unit in the various embodiments of the present application can be integrated in one processing unit, or each unit can exist physically as a separate unit, or two or more units can be integrated in one unit.

[0158] If the functions are implemented in the form of software function units and sold or used as independent products, they can be stored in a non-volatile computer readable storage medium executable by a processor. Based on this understanding, the technical solutions of the present application essentially or the part that contributes to the prior art, or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application. The foregoing storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes.

[0159] Finally, it should be noted that the above-described embodiments are merely specific embodiments of the present application, which are used to illustrate the technical solutions of the present application, but not to limit the present application, and the protection scope of the present application is not limited thereto. Although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still make modifications or easily think of changes to the technical solutions recorded in the foregoing embodiments, or make equivalent replacements to some technical features therein, within the technical range disclosed by the present application. The modifications, changes or replacements do not make the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for surface burst detection under a spaceborne multi-computing board architecture, characterized in that, The application is applied to a ground burst event detection device under a satellite-borne multi-computing board architecture, comprising a remote sensing data source module, a first computing module, a second computing module, a third computing module and a hybrid expert computing module. The method comprises: The remote sensing data source module sends remote sensing image data to the first computing module, the second computing module and the third computing module respectively; The first computing module, the second computing module and the third computing module respectively extract burst anomaly indicators from the remote sensing image data to obtain first burst anomaly indicators, second burst anomaly indicators and third burst anomaly indicators; wherein the first burst anomaly indicators refer to the burst anomaly indicators output by the first computing module, the second burst anomaly indicators refer to the burst anomaly indicators output by the second computing module, and the third burst anomaly indicators refer to the burst anomaly indicators output by the third computing module; the first burst anomaly indicators comprise at least one of a normalized vegetation index, a normalized difference water index and a difference vegetation index; the second burst anomaly indicators comprise a co-occurrence matrix and a binary code, which are used to describe spatial texture features; the third burst anomaly indicators comprise color, shape and semantic information of a ground burst event; The hybrid expert computing module determines the weighting coefficients corresponding to the first burst anomaly indicators, the second burst anomaly indicators and the third burst anomaly indicators according to the monitoring task corresponding to the remote sensing image data, and performs weighted fusion on the first burst anomaly indicators, the second burst anomaly indicators and the third burst anomaly indicators according to at least one fusion strategy to obtain at least one ground burst event detection result.

2. The method of claim 1, wherein, The number of remote sensing images is at least one frame; the first computing module, the second computing module and the third computing module respectively extract burst anomaly indicators from each remote sensing image data to obtain the first burst anomaly indicators, the second burst anomaly indicators and the third burst anomaly indicators corresponding to each remote sensing image data, comprising: For each frame of remote sensing image data, the first computing module extracts burst anomaly indicators from the remote sensing image data according to a remote sensing anomaly index calculation model to obtain first burst anomaly indicators, and sends interrupt information to the hybrid expert computing module; For each frame of remote sensing image data, the second computing module extracts burst anomaly indicators from the remote sensing image data according to a spatial texture feature extraction model to obtain second burst anomaly indicators, and sends interrupt information to the hybrid expert computing module; For each frame of remote sensing image data, the third computing module extracts burst anomaly indicators from the remote sensing image data according to a deep feature extraction model to obtain third burst anomaly indicators, and sends interrupt information to the hybrid expert computing module.

3. The method of claim 2, wherein, The hybrid expert computing module determines the weighting coefficients corresponding to the first burst anomaly indicators, the second burst anomaly indicators and the third burst anomaly indicators according to the monitoring task corresponding to the remote sensing image data, comprising: When the number of interrupt information received by the hybrid expert computing module is greater than a preset number, the first burst anomaly indicator, the second burst anomaly indicator and the third burst anomaly indicator are read; According to the monitoring task corresponding to the remote sensing image data, the weighting coefficients corresponding to the first burst anomaly indicator, the second burst anomaly indicator and the third burst anomaly indicator are queried.

4. The method of claim 1, wherein, The fusion strategy includes a weighted average strategy and a maximum voting strategy.

5. The method of claim 1, wherein, The hybrid expert computing module includes an adaptive learning model; The hybrid expert computing module is configured to obtain the first burst anomaly indicator, the second burst anomaly indicator, the third burst anomaly indicator, a first historical burst anomaly indicator, a second historical burst anomaly indicator and a third historical burst anomaly indicator; and train the adaptive learning model according to the first burst anomaly indicator, the second burst anomaly indicator, the third burst anomaly indicator, the first historical burst anomaly indicator, the second historical burst anomaly indicator and the third historical burst anomaly indicator, to obtain a trained adaptive learning model. The trained adaptive learning model is configured to determine the weighting coefficients corresponding to the first burst anomaly indicator, the second burst anomaly indicator and the third burst anomaly indicator according to the monitoring task corresponding to the remote sensing image data, and perform weighted fusion on the first burst anomaly indicator, the second burst anomaly indicator and the third burst anomaly indicator according to at least one fusion strategy, to obtain at least one surface burst event detection result.

6. The method of claim 2, wherein, The first computing module, the second computing module and the third computing module have the same structure, including an FPGA, a CPU, a GPU and a memory; the FPGA, the CPU and the GPU are connected in series, and the memory is connected to the FPGA and the CPU respectively; The FPGA is configured to obtain each remote sensing image data, sequentially send the remote sensing image data to the CPU in a preset order, and send the first burst anomaly indicator, the second burst anomaly indicator or the third burst anomaly indicator output by the CPU to the hybrid expert computing module; The CPU is configured to, for each frame of remote sensing image data, perform burst anomaly indicator extraction on the remote sensing image data, to obtain the first burst anomaly indicator, the second burst anomaly indicator or the third burst anomaly indicator; The GPU is configured to deploy the deep feature extraction model, so that the CPU performs burst anomaly indicator extraction on the remote sensing image data to obtain the third burst anomaly indicator; The memory is configured to, for each frame of remote sensing image data, store the remote sensing image data, and the first burst anomaly indicator, the second burst anomaly indicator or the third burst anomaly indicator.

7. The method of claim 6, wherein, The memory is also used to clear the remote sensing image data and the first burst anomaly index, the second burst anomaly index or the third burst anomaly index stored in the memory after the FPGA sends the first burst anomaly index, the second burst anomaly index or the third burst anomaly index to the hybrid expert computing module.

8. The method of claim 6, wherein, The CPU further comprises an external network module, configured to acquire the remote sensing image data through wireless communication.

9. An electronic device, comprising: The computer readable storage medium stores computer executable instructions, and the computer executable instructions, when invoked and executed by the processor, cause the processor to implement the satellite-borne multi-computing board architecture based ground burst event detection method in any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium stores computer executable instructions, and the computer executable instructions, when invoked and executed by the processor, cause the processor to implement the satellite-borne multi-computing board architecture based ground burst event detection method in any one of claims 1 to 8.

Images