Method and device for generating graph structure integrating anisotropy of space-time cube

Abstract

The present invention discloses a method and device for generating a graph structure integrating anisotropy of space-time cubes, which relates to the technical field of image processing and remote sensing data processing, and the main purpose is to solve the problem that the existing remote sensing data processing is affected by missing values, resulting in low accuracy. It mainly includes obtaining remote sensing data to be processed, and extracting multiple sub-space-time cubes based on the remote sensing data to be processed; extracting effective elements from the space-time cube to obtain a node set; constructing edges based on the effective nodes in the node set to obtain an edge set matching the node set, wherein the attributes of each edge in the edge set are embedded with the space-time offset of the corresponding effective node, and the space-time offset is used to characterize the space-time distance and anisotropy between the corresponding effective nodes; generating a graph structure based on the node set and the edge set matching the node set to obtain the graph structure of multiple sub-space-time cubes corresponding to the remote sensing data to be processed. It is mainly used to generate the graph structure of remote sensing data.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing and remote sensing data processing, and in particular to a method and device for generating a graph structure integrating anisotropy of space-time cubes. Background Art

[0002] Remote sensing data refers to data obtained by observing and acquiring targets from a long distance using remote sensing technology, which contains a large amount of image information and metadata. Remote sensing data has a wide range of application value and significance in many fields such as earth resource survey, vegetation classification, land use planning, crop pests and diseases, crop yield survey, environmental pollution monitoring, marine research and development, and earthquake monitoring. How to organize and process remote sensing data is the key to deeply analyze the temporal and spatial characteristics contained in remote sensing images and transform them into practical application results.

[0003] Existing remote sensing data processing methods mainly process remote sensing data based on two-dimensional or three-dimensional raster data structures. However, this method usually relies on simple spatiotemporal proximity or interpolation methods to fill in data missing areas, ignoring the impact of a large number of missing values ​​in the original image on the image's ability to express spatiotemporal relationships. This often causes image details to be missing and introduces uncertainty, resulting in significant deviations between the remote sensing data processing results and the actual situation, greatly reducing the accuracy of remote sensing data processing. Summary of the invention

[0004] In view of this, the present invention provides a method and device for generating a graph structure integrating the anisotropy of space-time cube, the main purpose of which is to solve the problem that the existing remote sensing data processing is affected by missing values, resulting in low accuracy.

[0005] According to one aspect of the present invention, a method for generating a graph structure integrating anisotropy of a space-time cube is provided, comprising:

[0006] Acquire remote sensing data to be processed, and extract multiple sub-space-time cubes based on the remote sensing data to be processed;

[0007] Extracting effective elements from each of the sub-space-time cubes to obtain a node set of the graph of each of the sub-space-time cubes;

[0008] Constructing edges according to valid nodes in the node set to obtain an edge set matching the node set, wherein the attribute of each edge in the edge set is embedded with a spatiotemporal offset of the corresponding valid node, and the spatiotemporal offset is used to characterize the spatiotemporal distance and anisotropy between the corresponding valid nodes;

[0009] A graph structure is generated based on the node set and the edge set matching the node set to obtain a graph structure of multiple sub-space-time cubes corresponding to the remote sensing data to be processed.

[0010] Furthermore, the edge construction is performed according to the valid nodes in the node set to obtain an edge set matching the node set, including:

[0011] For any of the valid nodes, determine a connection node corresponding to the valid node from the valid nodes included in the node set, and complete edge construction of the valid node based on the connection node;

[0012] For each edge, extract the spatiotemporal offset between corresponding valid nodes, and embed the spatiotemporal offset into the attribute of the edge to obtain an edge embedded with the spatiotemporal offset;

[0013] An edge set matching the node set is generated according to all the edges embedded with the time-space offsets.

[0014] Furthermore, for any of the valid nodes, determining a connection node corresponding to the valid node from the valid nodes included in the node set, and completing edge construction of the valid node based on the connection node includes:

[0015] For any valid node, perform the following operations to complete the edge construction of the valid node:

[0016] In the case where there are no other valid nodes on the connection path between the valid node and the central node, the central node is used as the connection node of the valid node, an edge between the valid node and the corresponding connection node is constructed, and a self-loop is added to the valid node;

[0017] When there are other valid nodes on the connection path between the valid node and the central node, the valid node closest to the valid node on the connection path is used as the connection node of the valid node, an edge between the valid node and the corresponding connection node is constructed, and a self-loop is added to the valid node.

[0018] Furthermore, extracting a plurality of sub-space-time cubes according to the remote sensing data to be processed includes:

[0019] Constructing a space-time cube based on the remote sensing data to be processed;

[0020] Constructing a buffer area of ​​the space-time cube according to a preset maximum expansion distance;

[0021] A plurality of central nodes are extracted in the non-buffer area, and the space-time cube is divided into a plurality of sub-space-time cubes with the central nodes as the three-dimensional geometric center.

[0022] Furthermore, the extracting effective elements from each of the sub-space-time cubes to obtain a node set of the graph of each of the sub-space-time cubes includes:

[0023] For each of the sub-space-time cubes, equal volume division is performed to obtain a plurality of space-time regions corresponding to the sub-space-time cube;

[0024] Polling each of the spatiotemporal regions, gradually extracting the elements in the spatiotemporal region that are closest to the central element as valid nodes, and embedding the pixel data of the valid nodes into the attributes of the valid nodes, until the number of valid nodes meets the hyperparameter of the preset number of nodes;

[0025] Extracting pixel data of the central node, and embedding the pixel data of the central node into the attributes of the central node accordingly;

[0026] A node set is generated according to the valid nodes embedded with pixel data and the central node, thereby obtaining node sets of different sub-space-time cubes.

[0027] Furthermore, extracting a plurality of sub-space-time cubes according to the remote sensing data to be processed includes:

[0028] Preprocessing the remote sensing data to be processed to obtain preprocessed remote sensing data, wherein the preprocessing includes at least one of radiation correction, atmospheric correction, geometric correction and noise elimination;

[0029] According to a high-quality pixel point extraction strategy, high-quality pixel points are extracted from the remote sensing data after the preprocessing, wherein the high-quality pixel point extraction strategy includes a spatial resolution identification strategy, a spectral resolution identification strategy, a radiation resolution identification strategy and a geometric accuracy identification strategy;

[0030] A space-time cube of the remote sensing data to be processed is constructed based on the high-quality pixel points and preset construction parameters.

[0031] Furthermore, after obtaining the graph structure of the plurality of sub-space-time cubes corresponding to the remote sensing data to be processed, the method further comprises:

[0032] Associating the graph structure with the remote sensing data to be processed and storing them in a target storage space;

[0033] In response to the analysis task of the remote sensing data to be processed, sending the graph structure of the remote sensing data to be processed to a target terminal, so that the target terminal performs the analysis task according to the graph structure;

[0034] The analysis tasks include a neighborhood average state acquisition task and a spatiotemporal density analysis task.

[0035] According to another aspect of the present invention, a device for generating a graph structure integrating anisotropy of a space-time cube is provided, comprising:

[0036] An acquisition module, used for acquiring remote sensing data to be processed, and extracting a plurality of sub-space-time cubes according to the remote sensing data to be processed;

[0037] A node extraction module, used for extracting effective elements from each of the sub-space-time cubes to obtain a node set of the graph of each of the sub-space-time cubes;

[0038] An edge construction module, configured to construct edges according to valid nodes in the node set to obtain an edge set matching the node set, wherein the attributes of each edge in the edge set are embedded with a spatiotemporal offset of the corresponding valid node, and the spatiotemporal offset is used to characterize the spatiotemporal distance and anisotropy between the corresponding valid nodes;

[0039] The graph structure generation module is used to generate a graph structure based on the node set and the edge set matching the node set, so as to obtain the graph structure of multiple sub-space-time cubes corresponding to the remote sensing data to be processed.

[0040] According to another aspect of the present invention, a storage medium is provided, wherein at least one executable instruction is stored in the storage medium, and the executable instruction enables a processor to execute operations corresponding to the above-mentioned method for generating a graph structure integrating anisotropy of space-time cube.

[0041] According to another aspect of the present invention, there is provided a terminal, comprising: a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other via the communication bus;

[0042] The memory is used to store at least one executable instruction, and the executable instruction enables the processor to execute operations corresponding to the above-mentioned graph structure generation method integrating the anisotropy of space-time cube.

[0043] By means of the above technical solution, the technical solution provided by the embodiment of the present invention has at least the following advantages:

[0044] The present invention provides a method and device for generating a graph structure integrating anisotropy of a space-time cube. The embodiment of the present invention obtains remote sensing data to be processed, and extracts multiple sub-space-time cubes according to the remote sensing data to be processed; extracts effective elements for each of the sub-space-time cubes to obtain a node set of the graph of each of the sub-space-time cubes; constructs edges according to the effective nodes in the node set to obtain an edge set matching the node set, wherein the attributes of each edge in the edge set are embedded with the space-time offset of the corresponding effective node, and the space-time offset is used to characterize the space-time distance and anisotropy between the corresponding effective nodes; generates a graph structure according to the node set and the edge set matching the node set to obtain the graph structure of multiple sub-space-time cubes corresponding to the remote sensing data to be processed. The anisotropy between each node in the remote sensing data space-time cube is fully considered, the influence of missing remote sensing data on the ability to express the space-time relationship of the image is taken into account, the probability of introducing uncertain data is reduced, and the consistency between the processed remote sensing data and the actual situation represented by the original remote sensing data is greatly improved, thereby greatly improving the accuracy of remote sensing data processing.

[0045] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present invention. Moreover, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings:

[0047] Figure 1 A flow chart of a method for generating a graph structure integrating anisotropy of a space-time cube provided by an embodiment of the present invention is shown;

[0048] Figure 2 A schematic diagram of space-time region division of a six-equal molecular space-time cube provided by an embodiment of the present invention is shown;

[0049] Figure 3 A schematic diagram of edge construction logic provided by an embodiment of the present invention is shown;

[0050] Figure 4 Another edge construction logic schematic diagram provided by an embodiment of the present invention is shown;

[0051] Figure 5A block diagram of a graph structure generation device integrating anisotropy of space-time cube provided by an embodiment of the present invention is shown;

[0052] Figure 6 A schematic diagram of the structure of a terminal provided by an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0053] The exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0054] In view of the problem that the existing remote sensing data processing is affected by missing values, resulting in low accuracy, the embodiment of the present invention provides a method for generating a graph structure integrating the anisotropy of space-time cubes, such as Figure 1 As shown, the method includes:

[0055] 101. Obtain remote sensing data to be processed, and extract multiple sub-space-time cubes based on the remote sensing data to be processed.

[0056] In the embodiment of the present invention, in order to extract the spatiotemporal features of the remote sensing data to be processed, the remote sensing data to be processed are arranged in the order of time, and a spatiotemporal cube is constructed according to the arrangement order. Among them, the spatiotemporal cube includes a pair of two-dimensional coordinate axes and a one-dimensional coordinate axis. The two-dimensional coordinate axis is used to represent the plane space of the real world, and the one-dimensional time axis is used to represent the change of the plane position along the time. The two-dimensional geometric position and the one-dimensional time constitute a three-dimensional cube model. In this spatiotemporal cube model, given any time point, the corresponding cross section, that is, the plane geometric state of the real world, can be obtained from the three-dimensional cube. Therefore, this model can intuitively express the time semantics. Among them, the remote sensing data to be processed can be any type of remote sensing data in any scene. For example, multispectral remote sensing images for feature recognition of vegetation, water bodies, etc., infrared remote sensing images for monitoring surface temperature distribution, passive microwave remote sensing data for monitoring ocean, atmosphere and surface humidity, etc. The embodiment of the present invention does not specifically limit the application scenario and type of the remote sensing data to be processed. In order to avoid the large amount of data in the space-time cube and the high performance requirements for computer data processing, the space-time cube is divided into multiple sub-space-time cubes, and corresponding graph structures are constructed for different sub-space-time cubes.

[0057] 102. Extract effective elements from each of the sub-space-time cubes to obtain a node set of the graph of each of the sub-space-time cubes.

[0058] In the embodiment of the present invention, after extracting multiple sub-space-time cubes, in order to construct the graph structure of the remote sensing data to be processed, it is necessary to extract the node sets corresponding to different sub-space-time cubes respectively, that is, the node set includes the node sets of different sub-space-time cubes. Then the construction of the graph structure is completed according to the nodes and edges. Therefore, the selection of nodes plays an important role in the construction of the graph structure. In order to ensure the quality of the subsequent graph structure construction, the sub-space-time cube is subjected to effective element extraction, that is, the nodes are screened to extract nodes with higher pixel quality and more representative in the remote sensing data as effective nodes, so as to form a node set based on these effective nodes. If the amount of remote sensing data is large and the constructed space-time cube is relatively large, in order to reduce the difficulty of data processing of the computer processor, it is necessary to divide the space-time cube into multiple sub-space-time cubes, and then extract effective elements for each sub-space-time cube to obtain the node set corresponding to each sub-space-time cube. The number of generated node sets corresponds to the number of sub-space-time cube segmentations, and the number of sub-space-time cube segmentations can be customized according to the application requirements of the actual scene, and the embodiment of the present invention does not make specific restrictions.

[0059] In one embodiment of the present invention, for further explanation and limitation, the extracting a plurality of sub-space-time cubes according to the remote sensing data to be processed includes:

[0060] Constructing a space-time cube based on the remote sensing data to be processed;

[0061] Constructing a buffer area of ​​the space-time cube according to a preset maximum expansion distance;

[0062] Extract multiple central nodes in the non-buffer area, and divide the space-time cube into multiple sub-space-time cubes with the central nodes as the three-dimensional geometric center;

[0063] Valid elements are extracted from each of the sub-space-time cubes to obtain a node set of the graph of each of the sub-space-time cubes.

[0064] In the embodiment of the present invention, in the process of dividing the sub-space-time cube, it is necessary to select a pixel point from the space-time cube as the central node of the sub-space-time cube, in order to avoid the situation where the central node is at the edge of the sub-space-time cube, resulting in the inability to divide the sub-space-time cube. Before dividing the sub-space-time cube, the buffer area of ​​the space-time cube is constructed with a preset maximum expansion distance. The nodes in this buffer area cannot be used as central nodes, that is, the central node cannot be selected from the buffer area. After the buffer area is constructed, in the non-buffer area, the corresponding number of central nodes are selected according to the expected number of sub-space-time cubes, and then the sub-space-time cubes are divided. Among them, the preset maximum expansion distance is the distance covering a sufficient number of pixels, and its value is not fixed. It varies according to the time and space intervals of the pixels. It can be customized according to the application scenario and data characteristics of the remote sensing data to be processed, and the example of the present invention does not make specific limitations.

[0065] In one embodiment of the present invention, for further explanation and limitation, the extracting effective elements from each of the sub-space-time cubes to obtain a node set of the graph of each of the sub-space-time cubes includes:

[0066] For each of the sub-space-time cubes, equal volume division is performed to obtain a plurality of space-time regions corresponding to the sub-space-time cube;

[0067] Polling each of the spatiotemporal regions, gradually extracting the elements in the spatiotemporal region that are closest to the central element as valid nodes, and embedding the pixel data of the valid nodes into the attributes of the valid nodes until the number of valid nodes meets the hyperparameter of the preset number of nodes;

[0068] Extracting pixel data corresponding to each of the valid nodes and the central node, and embedding the pixel data into the attributes of the valid nodes and the central node accordingly;

[0069] A node set is generated according to the valid nodes embedded with pixel data and the central node, thereby obtaining node sets of different sub-space-time cubes.

[0070] In the embodiment of the present invention, the effective node extraction process is performed for each sub-space-time cube. In order to improve the uniformity of the selected effective nodes in the space-time distribution, the sub-space-time cube is divided into equal volumes to divide the sub-space-time cube into multiple space-time regions. The polling method is used to cyclically visit each space-time region in turn. When visiting each region, the distance between each pixel point in the region and the central node (central element) of the sub-space-time cube is sorted (to ensure the dispersion of the node in the space-time distribution), and the pixel point closest to the central node in the region is extracted as a valid node until the number of valid nodes meets the node number hyperparameter of the preset complete graph, and the polling is stopped. After the effective node extraction is completed, the pixel data corresponding to the effective node is embedded in the attribute of the effective node to obtain the node set corresponding to each sub-space-time cube. Among them, the pixel data may include one or more of spectral value, brightness, grayscale, color, radiation correction coefficient, and geometric correction parameter, which varies according to the different acquisition scenes of the remote sensing image to be processed. For example, night remote sensing data only contains brightness, while daytime remote sensing data contains multiple data such as spectral value, brightness, color, etc.

[0071] It should be noted that the central node is also a valid node. In the process of pixel data embedding and node set generation, the processing of the central node and the valid node is the same. By dividing the sub-space-time cube into time-space regions, it is helpful to maintain the balance of the structure during the node selection process, and it is also more conducive to extracting representative pixels from different time-space regions.

[0072] It should be noted that the division method of the spatiotemporal region can be customized according to the node extraction accuracy requirements. For example, the sub-spatiotemporal cube is divided into 6 equal sub-spatiotemporal regions along the docking line, and divided into 8 equal sub-spatiotemporal regions along the center, etc., which is not specifically limited in the embodiment of the present invention. Figure 2 As shown in the figure, it is a schematic diagram of the division method of dividing the sub-space-time cube into 6 equal space-time regions along the docking line. In the figure, t is the time axis, w and h are the two vertical axes of the ground geometric plane presented by the remote sensing image, and the space-time regions of 1 to 6 quadrangular pyramids in the figure are obtained by dividing along the 45° diagonal line between each two axes.

[0073] 103. Construct edges according to valid nodes in the node set to obtain an edge set matching the node set.

[0074] In the embodiment of the present invention, after obtaining the valid nodes, another key element edge for constructing the graph structure is also required. In order to ensure the accuracy and consistency of the graph structure network, it is necessary to construct the edge according to the preset connection conditions. The connection conditions here can be customized according to the specific application requirements, and the embodiment of the present invention does not make specific limitations. The edge is a line or path connecting two valid nodes, representing the spatiotemporal connection between the valid nodes, which is conducive to a more intuitive understanding of the network structure and the spatiotemporal relationship between the valid nodes. It should be noted that the attributes of each edge in the edge set are embedded with the spatiotemporal offset of the corresponding valid node. Among them, the spatiotemporal offset can also be called the spatiotemporal position deviation, which can be expressed in the form of a multidimensional vector composed of time coordinates and two-dimensional space coordinates. It is used to characterize the spatiotemporal distance and anisotropy between the corresponding valid nodes. By embedding the spatiotemporal position deviation in the edge attribute, not only can the spatiotemporal distance and anisotropy between the valid nodes in the spatiotemporal cube be accurately described, but also a highly explanatory and accurate data reference can be provided when processing the missing value problem of remote sensing data based on the graph structure.

[0075] In one embodiment of the present invention, for further explanation and limitation, the step of constructing edges based on valid nodes in the node set to obtain an edge set matching the node set includes:

[0076] For any of the valid nodes, determine a connection node corresponding to the valid node from the valid nodes included in the node set, and complete edge construction of the valid node based on the connection node;

[0077] For each edge, extract the spatiotemporal offset between corresponding valid nodes, and embed the spatiotemporal offset into the attribute of the edge to obtain an edge embedded with the spatiotemporal offset;

[0078] An edge set matching the node set is generated according to all the edges embedded with the time-space offsets.

[0079] In an embodiment of the present invention, in order to realize the edge construction of a valid node, suitable connecting nodes are searched for the valid nodes one by one, and the line between the connecting node and the valid node is the edge of the valid node. Among them, the connecting node can be a central node or other valid nodes. After completing the edge construction, the spatiotemporal offset between the two valid nodes connected by the edge is extracted, that is, the vector composed of the offsets of the two valid nodes in each axial direction in the spatiotemporal cube coordinate system can be expressed as [T, W, H]. Among them, T is the deviation value of the two valid nodes in the time dimension, and W and H are the deviation values ​​of the two valid nodes in different directions in the two-dimensional geometric plane dimension. After obtaining the spatiotemporal offset, the spatiotemporal offset is embedded into the attribute of the corresponding edge so that the edge carries the anisotropy of the spatiotemporal cube.

[0080] It should be noted that since the spatiotemporal offset includes the correlation between the feature data of the same space at different time points, as well as the feature data of different spatial positions at the same time point, when the remote sensing data contains a large number of missing values, it can still accurately express the spatial continuity and temporal mutation of the remote sensing data, so that the generated graph structure can express the spatiotemporal characteristics that are closer to the actual situation, thereby ensuring the accuracy of remote sensing data processing.

[0081] In one embodiment of the present invention, for further explanation and limitation, determining a connection node corresponding to the valid node from the valid nodes included in the node set, and completing edge construction of the valid node based on the connection node includes:

[0082] For any valid node, perform the following operations to complete the edge construction of the valid node:

[0083] In the case where there are no other valid nodes on the connection path between the valid node and the central node, the central node is used as the connection node of the valid node, an edge between the valid node and the corresponding connection node is constructed, and a self-loop is added to the valid node;

[0084] When there are other valid nodes on the connection path between the valid node and the central node, the valid node closest to the valid node on the connection path is used as the connection node of the valid node, an edge between the valid node and the corresponding connection node is constructed, and a self-loop is added to the valid node.

[0085] In an embodiment of the present invention, during the edge construction process, the edges of the valid nodes are constructed in sequence in ascending order of the distance between the valid node and the central node. When the connection path between the valid node and the central node does not pass through other valid nodes, or the passed nodes are invalid nodes or low-quality pixels, the valid node and the central node are directly connected, that is, the central node is used as the connection node, and the edge is constructed with the valid node. When the connection path between the valid node and the central node passes through other valid nodes, it indicates that the current valid node is a node far away from the central node, and the other valid nodes passed through are valid nodes that have completed edge construction first. Therefore, the one of the valid nodes that is closest to the current valid node is used as the connection node, and the edge is constructed with the valid node. The calculation of the connection path can be implemented using the Bresenham algorithm. After completing the edge construction between each valid node and other valid nodes, a self-loop is added to the valid node, that is, an information transmission path inside the node is constructed. By adding a self-loop to the valid node, it can be ensured that the information of the valid node itself is not lost. Figure 3 , 4As shown in the figure, "?" indicates the central node, the solid line box farthest from the central node is the current valid node that needs to be constructed, the dashed line box is an invalid or low-quality pixel, the dashed line is the connection between the central node and the current valid node, and the solid line is an existing edge. Figure 3 In the two cases shown in the figure, an edge as shown by the dashed line is constructed between the current valid node and the central node, and a self-loop is added to the valid node. Figure 4 In the case shown, the connection path between the current valid node and the central node passes through two other valid nodes, and the valid node represented by the solid line box adjacent to the right side of the current valid node is taken as the connection node.

[0086] 104. Generate a graph structure based on the node set and the edge set matching the node set to obtain a graph structure of multiple sub-space-time cubes corresponding to the remote sensing data to be processed.

[0087] In the embodiment of the present invention, when the node set includes multiple nodes, it is necessary to construct a graph structure according to each set of node combinations and edge sets respectively, so as to obtain graph structures corresponding to different node sets, that is, the remote sensing data to be processed can be represented in the form of multiple graph structures. The generation of the graph structure can be achieved through an open source graph structure generation tool, or through a self-developed graph structure generation algorithm or template, which is not specifically limited in the embodiment of the present invention.

[0088] It should be noted that since the edges used to construct the graph structure are embedded with the spatiotemporal offsets used to characterize the spatiotemporal distances and anisotropy between the corresponding valid nodes, the graph structure can fully express the spatiotemporal anisotropy contained in the time series remote sensing data containing missing values, thereby achieving the purpose of comprehensively characterizing the complex relationship of spatiotemporal data. At the same time, in the process of graph construction, there is no need to consider the existence of missing values, nor will it be affected by missing values, thus avoiding the error influence of missing value filling in remote sensing data, thereby ensuring the accuracy of remote sensing data processing.

[0089] In one embodiment of the present invention, for further explanation and limitation, after obtaining the graph structure of multiple sub-space-time cubes corresponding to the remote sensing data to be processed, the method further includes:

[0090] Associating the graph structure with the remote sensing data to be processed and storing them in a target storage space;

[0091] In response to the analysis task of the remote sensing data to be processed, the graph structure of the remote sensing data to be processed is sent to a target terminal, so that the target terminal performs the analysis task according to the graph structure.

[0092] In an embodiment of the present invention, the generated graph structure can be used as a data basis and data basis in a variety of subsequent remote sensing data processing tasks. After the graph structure is generated, the remote sensing data to be processed is associated with the graph structure, and the graph structure is stored in a pre-constructed target storage space in a form that can be understood and stored by a computer, so that the graph structure can be retrieved in subsequent remote sensing data processing tasks based on the graph structure. Among them, the target storage space can be local storage, cloud storage or blockchain storage, and the embodiment of the present invention does not make specific limitations. When an analysis task for processing remote sensing data to be processed is received, the graph structure of the remote sensing data to be processed is retrieved and sent to a target terminal for performing the analysis task, so that the target terminal can perform further remote sensing data processing and analysis based on the graph structure. Among them, the analysis task includes a neighborhood average state acquisition task and a spatiotemporal density analysis task, that is, preferably, the graph structure can be applied to the remote sensing data processing tasks of neighborhood average state acquisition and spatiotemporal density analysis. Since the graph structure is built based on edges embedded with spatiotemporal offsets, it can characterize spatiotemporal anisotropy. In the process of further remote sensing data based on the graph structure, it can take into account spatiotemporal specificity and provide highly interpretable and accurate representation data when dealing with the problem of missing values ​​in remote sensing data, thereby improving the accuracy of remote sensing data processing.

[0093] In one embodiment of the present invention, for further explanation and limitation, the extracting a plurality of sub-space-time cubes according to the remote sensing data to be processed includes:

[0094] Preprocessing the remote sensing data to be processed to obtain remote sensing data after preprocessing;

[0095] Extracting high-quality pixel points from the remote sensing data after the preprocessing according to a high-quality pixel point extraction strategy;

[0096] A space-time cube of the remote sensing data to be processed is constructed based on the high-quality pixel points and preset construction parameters.

[0097] In the embodiment of the present invention, in order to ensure the quality of the nodes used to construct the graph structure, before constructing the space-time cube, the remote sensing data to be processed is preprocessed to filter the noise in the original remote sensing data, and correct the image distortion caused by environmental factors such as radiation and atmosphere. Specifically, the content of the preprocessing includes at least one of radiation correction, atmospheric correction, geometric correction and noise elimination. That is, according to the acquisition environment of the remote sensing data to be processed, one, two or more preprocessing methods of radiation correction, atmospheric correction, geometric correction and noise elimination can be selected to process the remote sensing data to be processed, and the embodiment of the present invention does not make specific restrictions. After completing the preprocessing, high-quality pixels are extracted according to the preset high-quality pixel extraction strategy. Among them, the high-quality pixel extraction strategy includes a spatial resolution recognition strategy, a spectral resolution recognition strategy, a radiation resolution recognition strategy and a geometric accuracy recognition strategy, and each strategy contains a corresponding recognition threshold. In the process of extracting according to the high-quality pixel extraction strategy, each pixel is traversed, and for each pixel, it is identified whether the corresponding parameter of this pixel satisfies the corresponding recognition threshold. If all parameters are better than the corresponding recognition threshold, it is selected as a high-quality pixel. For example, if the spatial resolution, spectral resolution, spatial resolution, radiation resolution, and geometric accuracy of a certain pixel point are all higher than the corresponding recognition threshold, the current pixel point is determined to be a high-quality pixel point. Of course, if after traversing all the pixels, the high-quality pixels obtained do not meet the preset number of pixels, the number requirements of the recognition strategy can be adjusted, such as pixels that only do not meet one recognition strategy are also selected as high-quality pixels, thereby obtaining a preset number of high-quality pixels. After obtaining high-quality pixels, an open source or self-developed space-time cube construction tool is used to construct a space-time cube based on the high-quality pixels and preset construction parameters. Among them, the preset construction parameters include parameters such as bars, time slices, and bar time series, which can be customized according to specific processing needs, and the embodiments of the present invention do not make specific limitations.

[0098] The present invention provides a method for generating a graph structure integrating the anisotropy of a space-time cube. The embodiment of the present invention obtains remote sensing data to be processed, and extracts multiple sub-space-time cubes according to the remote sensing data to be processed; extracts effective elements for each of the sub-space-time cubes to obtain a node set of the graph of each of the sub-space-time cubes; constructs edges according to the effective nodes in the node set to obtain an edge set matching the node set, wherein the attributes of each edge in the edge set are embedded with the space-time offset of the corresponding effective node, and the space-time offset is used to characterize the space-time distance and anisotropy between the corresponding effective nodes; generates a graph structure according to the node set and the edge set matching the node set to obtain the graph structure of multiple sub-space-time cubes corresponding to the remote sensing data to be processed. The anisotropy between each node in the space-time cube of remote sensing data is fully considered, the influence of missing remote sensing data on the ability to express the space-time relationship of images is taken into account, the probability of introducing uncertain data is reduced, and the consistency between the processed remote sensing data and the actual situation represented by the original remote sensing data is greatly improved, thereby greatly improving the accuracy of remote sensing data processing.

[0099] Furthermore, as a response to the above Figure 1 The implementation of the method shown in the figure, the embodiment of the present invention provides a graph structure generation device integrating the anisotropy of space-time cube, such as Figure 5 As shown, the device comprises:

[0100] An acquisition module 31 is used to acquire the remote sensing data to be processed, and extract a plurality of sub-space-time cubes according to the remote sensing data to be processed;

[0101] A node extraction module 32 is used to extract effective elements from each of the sub-space-time cubes to obtain a node set of the graph of each of the sub-space-time cubes;

[0102] An edge construction module 33, configured to construct edges according to valid nodes in the node set to obtain an edge set matching the node set, wherein the attributes of each edge in the edge set are embedded with a spatiotemporal offset of the corresponding valid node, and the spatiotemporal offset is used to characterize the spatiotemporal distance and anisotropy between the corresponding valid nodes;

[0103] The graph structure generation module 34 is used to generate a graph structure according to the node set and the edge set matching the node set, so as to obtain the graph structure of multiple sub-space-time cubes corresponding to the remote sensing data to be processed.

[0104] Furthermore, the edge construction module 33 includes:

[0105] An edge construction unit, configured to determine, for any of the valid nodes, a connection node corresponding to the valid node from the valid nodes included in the node set, and complete edge construction of the valid node based on the connection node;

[0106] An embedding unit, used for extracting the spatiotemporal offset between corresponding valid nodes for each edge, and embedding the spatiotemporal offset into the attribute of the edge to obtain the edge embedded with the spatiotemporal offset;

[0107] A generating unit is used to generate an edge set matching the node set according to all the edges embedded with the spatiotemporal offsets.

[0108] Further, in a specific application scenario, the edge construction unit is specifically used to perform the following operations for any of the valid nodes to complete the edge construction of the valid node:

[0109] In the case where there are no other valid nodes on the connection path between the valid node and the central node, the central node is used as the connection node of the valid node, an edge between the valid node and the corresponding connection node is constructed, and a self-loop is added to the valid node;

[0110] When there are other valid nodes on the connection path between the valid node and the central node, the valid node closest to the valid node on the connection path is used as the connection node of the valid node, an edge between the valid node and the corresponding connection node is constructed, and a self-loop is added to the valid node.

[0111] Furthermore, the node extraction module 32 includes:

[0112] A buffer construction unit, used to construct a buffer area of ​​the space-time cube according to a preset maximum expansion distance;

[0113] A segmentation unit, used for extracting a plurality of central nodes in the non-buffer area, and taking the central nodes as the three-dimensional geometric center, segmenting the space-time cube into a plurality of sub-space-time cubes;

[0114] The first node extraction unit is used to extract effective elements from each of the sub-space-time cubes to obtain a node set of the graph of each of the sub-space-time cubes.

[0115] Further, in a specific application scenario, the first node extraction unit is specifically used to perform equal volume division on each of the sub-space-time cubes to obtain a plurality of space-time regions corresponding to the sub-space-time cubes;

[0116] Polling each of the spatiotemporal regions, gradually extracting the elements in the spatiotemporal region that are closest to the central element as valid nodes, and embedding the pixel data of the valid nodes into the attributes of the valid nodes until the number of valid nodes meets the hyperparameter of the preset number of nodes;

[0117] Extracting pixel data corresponding to each of the valid nodes and the central node, and embedding the pixel data into the attributes of the valid nodes and the central node accordingly;

[0118] A node set is generated according to the valid nodes embedded with pixel data and the central node, thereby obtaining node sets of different sub-space-time cubes.

[0119] Furthermore, the acquisition module 31 includes:

[0120] A preprocessing unit, used for preprocessing the remote sensing data to be processed to obtain remote sensing data after preprocessing, wherein the preprocessing includes at least one of radiation correction, atmospheric correction, geometric correction and noise elimination;

[0121] A second node extraction unit is used to extract high-quality pixels from the remote sensing data after the preprocessing according to a high-quality pixel extraction strategy, wherein the high-quality pixel extraction strategy includes a spatial resolution recognition strategy, a spectral resolution recognition strategy, a radiation resolution recognition strategy and a geometric accuracy recognition strategy;

[0122] The space-time cube construction unit is used to construct the space-time cube of the remote sensing data to be processed according to the high-quality pixel points and preset construction parameters.

[0123] Furthermore, the device also includes:

[0124] A storage module, used for associating the graph structure with the remote sensing data to be processed and storing them in a target storage space;

[0125] A sending module is used to send the graph structure of the remote sensing data to be processed to a target terminal in response to the analysis task of the remote sensing data to be processed, so that the target terminal performs the analysis task according to the graph structure; wherein the analysis task includes a neighborhood average state acquisition task and a spatiotemporal density analysis task.

[0126] The present invention provides a graph structure generation device integrating the anisotropy of space-time cube. The embodiment of the present invention obtains the remote sensing data to be processed, and extracts multiple sub-space-time cubes according to the remote sensing data to be processed; extracts effective elements for each of the sub-space-time cubes to obtain the node set of the graph of each of the sub-space-time cubes; constructs edges according to the effective nodes in the node set to obtain the edge set matching the node set, wherein the attributes of each edge in the edge set are embedded with the space-time offset of the corresponding effective node, and the space-time offset is used to characterize the space-time distance and anisotropy between the corresponding effective nodes; generates the graph structure according to the node set and the edge set matching the node set to obtain the graph structure of multiple sub-space-time cubes corresponding to the remote sensing data to be processed. The anisotropy between each node in the remote sensing data space-time cube is fully considered, the influence of the missing remote sensing data on the ability to express the space-time relationship of the image is taken into account, the probability of introducing uncertain data is reduced, and the consistency between the processed remote sensing data and the actual situation represented by the original remote sensing data is greatly improved, thereby greatly improving the accuracy of remote sensing data processing.

[0127] According to an embodiment of the present invention, a storage medium is provided, wherein the storage medium stores at least one executable instruction, and the computer executable instruction can execute the method for generating a graph structure integrating the anisotropy of the space-time cube in any of the above method embodiments.

[0128] Figure 6 A schematic diagram of the structure of a terminal provided according to an embodiment of the present invention is shown. The specific embodiment of the present invention does not limit the specific implementation of the terminal.

[0129] like Figure 6 As shown, the terminal may include: a processor (processor) 402 , a communication interface (Communications Interface) 404 , a memory (memory) 406 , and a communication bus 408 .

[0130] The processor 402 , the communication interface 404 , and the memory 406 communicate with each other via a communication bus 408 .

[0131] The communication interface 404 is used for network communication with other devices such as a client or other servers.

[0132] The processor 402 is used to execute the program 410, and specifically can execute the relevant steps in the above-mentioned embodiment of the method for generating a graph structure integrating the anisotropy of the space-time cube.

[0133] Specifically, the program 410 may include program codes, which include computer operation instructions.

[0134] The processor 402 may be a central processing unit (CPU), or an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiment of the present invention. The one or more processors included in the terminal may be processors of the same type, such as one or more CPUs; or processors of different types, such as one or more CPUs and one or more ASICs.

[0135] The memory 406 is used to store the program 410. The memory 406 may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory.

[0136] The program 410 may be specifically configured to enable the processor 402 to perform the following operations:

[0137] Acquire remote sensing data to be processed, and extract multiple sub-space-time cubes based on the remote sensing data to be processed;

[0138] Extracting effective elements from each of the sub-space-time cubes to obtain a node set of the graph of each of the sub-space-time cubes;

[0139] Constructing edges according to valid nodes in the node set to obtain an edge set matching the node set, wherein the attribute of each edge in the edge set is embedded with a spatiotemporal offset of the corresponding valid node, and the spatiotemporal offset is used to characterize the spatiotemporal distance and anisotropy between the corresponding valid nodes;

[0140] A graph structure is generated based on the node set and the edge set matching the node set to obtain a graph structure of multiple sub-space-time cubes corresponding to the remote sensing data to be processed.

[0141] Obviously, those skilled in the art should understand that the above modules or steps of the present invention can be implemented by a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, and optionally, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, and in some cases, the steps shown or described can be executed in a different order than here, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present invention is not limited to any specific combination of hardware and software.

[0142] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A graph structure generation method integrating space-time cube anisotropy, characterized in that: include: Acquire remote sensing data to be processed, and extract multiple sub-space-time cubes based on the remote sensing data to be processed; Extracting effective elements from each of the sub-space-time cubes to obtain a node set of the graph of each of the sub-space-time cubes; The method comprises: for any valid node, determining a connection node corresponding to the valid node among the valid nodes included in the node set, and completing the edge construction of the valid node based on the connection node, specifically comprising: for any valid node, performing the following operations to complete the edge construction of the valid node: when there are no other valid nodes on the connection path between the valid node and the central node, taking the central node as the connection node of the valid node, constructing an edge between the valid node and the corresponding connection node, and adding a self-loop for the valid node; In the case where there are other valid nodes on the connection path between the valid node and the central node, the valid node closest to the valid node on the connection path is used as the connection node of the valid node, an edge between the valid node and the corresponding connection node is constructed, and a self-loop is added to the valid node; For each edge, extract the spatiotemporal offset between corresponding valid nodes, and embed the spatiotemporal offset into the attribute of the edge to obtain an edge embedded with the spatiotemporal offset; generate an edge set matching the node set based on all the edges embedded with the spatiotemporal offset, wherein the attribute of each edge in the edge set is embedded with the spatiotemporal offset of the corresponding valid node, and the spatiotemporal offset is used to characterize the spatiotemporal distance and anisotropy between the corresponding valid nodes; A graph structure is generated based on the node set and the edge set matching the node set to obtain a graph structure of multiple sub-space-time cubes corresponding to the remote sensing data to be processed.

2. The method according to claim 1, characterized in that The step of extracting a plurality of sub-space-time cubes according to the remote sensing data to be processed comprises: Constructing a space-time cube based on the remote sensing data to be processed; Constructing a buffer area of ​​the space-time cube according to a preset maximum expansion distance; A plurality of central nodes are extracted in the non-buffer area, and the space-time cube is divided into a plurality of sub-space-time cubes with the central nodes as the three-dimensional geometric center.

3. The method according to claim 2, characterized in that The extracting effective elements from each of the sub-space-time cubes to obtain a node set of the graph of each of the sub-space-time cubes includes: For each of the sub-space-time cubes, equal volume division is performed to obtain a plurality of space-time regions corresponding to the sub-space-time cube; Polling each of the spatiotemporal regions, gradually extracting the elements in the spatiotemporal region that are closest to the central element as valid nodes, and embedding the pixel data of the valid nodes into the attributes of the valid nodes until the number of valid nodes meets the hyperparameter of the preset number of nodes; Extracting pixel data of the central node, and embedding the pixel data of the central node into the attributes of the central node; A node set is generated according to the valid nodes embedded with pixel data and the central node, thereby obtaining node sets of different sub-space-time cubes.

4. The method according to claim 2, characterized in that: The step of constructing a space-time cube based on the remote sensing data to be processed includes: Preprocessing the remote sensing data to be processed to obtain preprocessed remote sensing data, wherein the preprocessing includes at least one of radiation correction, atmospheric correction, geometric correction and noise elimination; According to a high-quality pixel point extraction strategy, high-quality pixel points are extracted from the remote sensing data after the preprocessing, wherein the high-quality pixel point extraction strategy includes a spatial resolution identification strategy, a spectral resolution identification strategy, a radiation resolution identification strategy and a geometric accuracy identification strategy; A space-time cube of the remote sensing data to be processed is constructed based on the high-quality pixel points and preset construction parameters.

5. The method according to claim 1, characterized in that After obtaining the graph structures of the plurality of sub-space-time cubes corresponding to the remote sensing data to be processed, the method further comprises: Associating the graph structure with the remote sensing data to be processed and storing them in a target storage space; In response to the analysis task of the remote sensing data to be processed, sending the graph structure of the remote sensing data to be processed to a target terminal, so that the target terminal performs the analysis task according to the graph structure; The analysis tasks include a neighborhood average state acquisition task and a spatiotemporal density analysis task.

6. A graph structure generation device integrating space-time cube anisotropy, characterized in that: include: An acquisition module, used for acquiring remote sensing data to be processed, and extracting a plurality of sub-space-time cubes according to the remote sensing data to be processed; A node extraction module, used for extracting effective elements from each of the sub-space-time cubes to obtain a node set of the graph of each of the sub-space-time cubes; An edge construction module, configured to construct edges according to valid nodes in the node set to obtain an edge set matching the node set, wherein the attributes of each edge in the edge set are embedded with a spatiotemporal offset of the corresponding valid node, and the spatiotemporal offset is used to characterize the spatiotemporal distance and anisotropy between the corresponding valid nodes; A graph structure generation module, used to generate a graph structure according to the node set and the edge set matching the node set, to obtain the graph structure of multiple sub-space-time cubes corresponding to the remote sensing data to be processed; The edge building module comprises: An edge construction unit is used to determine, for any of the valid nodes, a connection node corresponding to the valid node from the valid nodes included in the node set, and complete edge construction of the valid node based on the connection node; in a specific application scenario, the edge construction unit is specifically used to perform the following operations for any of the valid nodes to complete edge construction of the valid node: In the case where there are no other valid nodes on the connection path between the valid node and the central node, the central node is used as the connection node of the valid node, an edge between the valid node and the corresponding connection node is constructed, and a self-loop is added to the valid node; In the case where there are other valid nodes on the connection path between the valid node and the central node, the valid node closest to the valid node on the connection path is used as the connection node of the valid node, an edge between the valid node and the corresponding connection node is constructed, and a self-loop is added to the valid node; An embedding unit, used for extracting the spatiotemporal offset between corresponding valid nodes for each edge, and embedding the spatiotemporal offset into the attribute of the edge to obtain the edge embedded with the spatiotemporal offset; A generating unit is used to generate an edge set matching the node set according to all the edges embedded with the spatiotemporal offsets.

7. A storage medium storing at least one executable instruction, wherein the executable instruction enables a processor to execute operations corresponding to the method for generating a graph structure of anisotropic fused space-time cubes as described in any one of claims 1 to 5.

8. A terminal, comprising: A processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other via the communication bus; The memory is used to store at least one executable instruction, and the executable instruction enables the processor to perform operations corresponding to the method for generating a graph structure of anisotropic fused space-time cubes according to any one of claims 1 to 5.

Images