UAV airborne radar data analysis method and system

Abstract

The present application discloses a method and system for analyzing radar data onboard a drone, and belongs to the field of data analysis technology. The present application deeply mines the multi-level and multi-dimensional semantic information of radar scanning data in the target spatiotemporal knowledge relationship space, and realizes the integrated processing of these semantic information, thereby obtaining a more comprehensive and in-depth data representation. Furthermore, the present application also realizes effective interaction with the environmental monitoring platform server, and by calculating the target matching probability, high-quality radar scanning data is promptly sent to the required server. In this way, it can provide more accurate and timely data support for environmental monitoring and governance, and promote further development in this field.

Description

Technical Field

[0001] The present application belongs to the field of data analysis technology, and specifically relates to a method and system for analyzing UAV airborne radar data. Background Art

[0002] In the field of environmental monitoring and governance, airborne radar 3D scanning data, as an important data source, can provide rich and detailed surface environmental information. However, traditional radar data analysis methods often only focus on the direct features of the scanning data, while ignoring the semantic information and spatiotemporal background behind it, resulting in limited depth and breadth of data analysis. At the same time, since radar scanning data is usually massive and complex, how to effectively match this data with specific environmental monitoring tasks is also a technical problem currently faced. Summary of the invention

[0003] The present application provides a method and system for analyzing drone-borne radar data, which can solve or partially solve the technical problems involved in the above-mentioned background technology.

[0004] The embodiment of the present application provides a method for analyzing airborne radar data of a drone, which is applied to a radar data analysis system. The method comprises: obtaining a target knowledge vector entity in a target spatiotemporal knowledge relationship space of airborne radar three-dimensional scanning data to be processed and a target association vector entity corresponding to the target knowledge vector entity, wherein the target association vector entity comprises a target linkage entity corresponding to target linkage radar scanning data involved in the airborne radar three-dimensional scanning data to be processed and a target area element entity corresponding to a surface area element of the airborne radar three-dimensional scanning data to be processed; performing multimodal semantic mining on the target knowledge vector entity and the target association vector entity to obtain an original multimodal entity semantic corresponding to the target knowledge vector entity, the target linkage entity entity and the target association vector entity. The target multimode linkage semantics corresponding to the body and the target area element semantics corresponding to the target area element entity; the original multimode entity semantics, the target multimode linkage semantics and the target area element semantics are integrated to obtain the target multimode entity semantics of the airborne radar three-dimensional scanning data to be processed; the multimode entity semantics corresponding to the environmental monitoring platform server in the target spatiotemporal knowledge relationship space are obtained; based on the target multimode entity semantics and the multimode entity semantics corresponding to the environmental monitoring platform server, the target matching probability of the environmental monitoring platform server for the airborne radar three-dimensional scanning data to be processed is determined; the airborne radar three-dimensional scanning data to be processed whose target matching probability is greater than the set probability is sent to the environmental monitoring platform server.

[0005] An embodiment of the present application provides a radar data analysis system, comprising at least one processor and a memory; the memory stores computer-executable instructions; the at least one processor executes the computer-executable instructions stored in the memory, so that the at least one processor performs the above method.

[0006] An embodiment of the present application provides a readable storage medium, on which a program or instruction is stored. When the program or instruction is executed by a processor, the steps of the above method are implemented.

[0007] The embodiment of the present application deeply mines the multi-level and multi-dimensional semantic information of radar scanning data in the target spatiotemporal knowledge relationship space, and realizes the integrated processing of these semantic information, thereby obtaining a more comprehensive and in-depth data representation. Furthermore, the embodiment of the present application also realizes effective interaction with the environmental monitoring platform server, and by calculating the target matching probability, high-quality radar scanning data is promptly sent to the required server. In this way, it can provide more accurate and timely data support for environmental monitoring and governance, and promote further development in this field. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] Figure 1 A flowchart of a method for analyzing drone-borne radar data provided in an embodiment of the present application.

[0009] Figure 2 A schematic diagram of the structure of a radar data analysis system provided in an embodiment of the present application. DETAILED DESCRIPTION

[0010] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0011] The terms "first", "second", etc. in this application are used to distinguish similar objects, and are not used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable under appropriate circumstances, so that the embodiments of the present application can be implemented in an order other than those illustrated or described here, and the objects distinguished by "first", "second", etc. are usually a class, and the number of objects is not limited. For example, the first object can be one or more. In addition, in this application, "and / or" represents at least one of the connected objects, and the character " / " generally indicates that the objects associated with each other are in an "or" relationship.

[0012] Figure 1A method for analyzing radar data onboard a drone is shown, which is applied to a radar data analysis system. The method includes the following steps 110 to 140.

[0013] Step 110, the radar data analysis system obtains a target knowledge vector entity in the target spatiotemporal knowledge relationship space of the airborne radar three-dimensional scanning data to be processed and a target association vector entity corresponding to the target knowledge vector entity, wherein the target association vector entity includes a target linkage entity corresponding to the target linkage radar scanning data involved in the airborne radar three-dimensional scanning data to be processed and a target area element entity corresponding to the surface area element of the airborne radar three-dimensional scanning data to be processed.

[0014] Step 120: The radar data analysis system performs multimodal semantic mining on the target knowledge vector entity and the target association vector entity to obtain the original multimodal entity semantics corresponding to the target knowledge vector entity, the target multimodal linkage semantics corresponding to the target linkage entity, and the target area element semantics corresponding to the target area element entity.

[0015] Step 130: The radar data analysis system integrates the original multi-mode entity semantics, the target multi-mode linkage semantics and the target area element semantics to obtain the target multi-mode entity semantics of the airborne radar three-dimensional scanning data to be processed.

[0016] Step 140: The radar data analysis system obtains the multimodal entity semantics corresponding to the environmental monitoring platform server in the target spatiotemporal knowledge relationship space; determines the target matching probability of the environmental monitoring platform server for the airborne radar three-dimensional scanning data to be processed based on the target multimodal entity semantics and the multimodal entity semantics corresponding to the environmental monitoring platform server; and sends the airborne radar three-dimensional scanning data to be processed whose target matching probability is greater than the set probability to the environmental monitoring platform server.

[0017] It can be understood that the embodiment of the present application provides a new perspective and accuracy for environmental monitoring through the introduction of drone technology, especially the use of radar equipment for three-dimensional scanning. However, how to efficiently process and analyze these massive and complex data has become a problem that needs to be solved urgently. Based on this, the above-mentioned radar data analysis system can deeply mine and analyze the airborne radar three-dimensional scanning data in the target spatiotemporal knowledge relationship space, thereby extracting valuable environmental information. The following will introduce the implementation process of the technical solution (step 110-step 140) in detail through a specific application scenario.

[0018] First, consider an actual environmental monitoring scenario: a certain region has faced serious water pollution and ecological damage in recent years. In order to conduct accurate environmental monitoring and governance in the region, relevant units introduced drones equipped with radar equipment for regular three-dimensional scanning. These scanned data contain rich environmental information such as surface morphology, water conditions, and vegetation distribution. However, how to extract valuable information from these massive data has become a huge challenge.

[0019] Based on this, the radar data analysis system first obtains the airborne radar 3D scanning data to be processed. In the target spatiotemporal knowledge relationship space, these data are parsed into target knowledge vector entities. Each entity represents a specific radar scanning data block, including its unique spatial position, timestamp, scanning parameters and other information.

[0020] At the same time, the radar data analysis system also identifies target-related vector entities involved in these scan data. These related vector entities include two aspects: one is other radar scan data that are temporally and spatially related to the airborne radar 3D scan data to be processed, namely, target linkage entities; the other is the surface area elements corresponding to the airborne radar 3D scan data to be processed, such as water bodies, vegetation, buildings, etc., namely, target area element entities.

[0021] Next, the radar data analysis system performs multimodal semantic mining on these target knowledge vector entities and target association vector entities. The purpose of this step is to extract deep semantic information from the data for subsequent analysis and application. The radar data analysis system first analyzes the characteristics and attributes of the target knowledge vector entity itself and obtains its corresponding original multimodal entity semantics. These semantic information include the morphological characteristics, spatial distribution, and temporal changes of the scanned data.

[0022] At the same time, the radar data analysis system also conducts in-depth analysis of the target linkage entity. By comparing the scan data at different time points, the radar data analysis system can capture the dynamic changes of elements such as surface morphology and water conditions. These change information is extracted as target multi-mode linkage semantics, which reflects the evolution of the surface environment over time.

[0023] In addition, the radar data analysis system also performs semantic analysis on the target area element entities. By identifying and analyzing the surface elements in the scanned data, the radar data analysis system can extract the semantics of the target area elements that correspond to them. These semantic information include the pollution level of water bodies, the coverage and distribution of vegetation, the structure and layout of buildings, etc.

[0024] After completing the multi-mode semantic mining, the radar data analysis system further integrates these semantic information. The purpose of this step is to fuse and integrate the original multi-mode entity semantics, the target multi-mode linkage semantics and the target area element semantics, so as to obtain the target multi-mode entity semantics of the airborne radar 3D scanning data to be processed. This semantic information reflects the overall status and characteristics of the surface environment more comprehensively and deeply.

[0025] Finally, the radar data analysis system connected and applied the target multimodal entity semantics to the environmental monitoring platform server. The radar data analysis system first obtained the corresponding multimodal entity semantics of the environmental monitoring platform server in the target spatiotemporal knowledge relationship space. These semantic information represent the existing environmental monitoring data and knowledge base of the environmental monitoring platform server.

[0026] Then, the radar data analysis system matches and compares the multimodal entity semantics of the target and the multimodal entity semantics corresponding to the environmental monitoring platform server. By calculating the similarity and correlation between the two, the radar data analysis system can determine the target matching probability of the environmental monitoring platform server for the airborne radar 3D scanning data to be processed. This probability reflects the matching degree and correlation between the data to be processed and the existing data of the environmental monitoring platform server.

[0027] When the target matching probability is greater than the set probability, it means that the airborne radar 3D scanning data to be processed has a high degree of matching and correlation with the existing data of the environmental monitoring platform server. At this time, the radar data analysis system sends this data to the environmental monitoring platform server for further environmental quality assessment, pollution warning and emergency response decision-making.

[0028] Through the implementation of this technical solution, the radar data analysis system successfully transforms drone radar scanning data into valuable environmental information, providing strong support for environmental monitoring and governance. First, the radar data analysis system can efficiently process and analyze massive and complex radar scanning data and extract deep semantic information from it; second, the radar data analysis system can connect and apply this semantic information to the environmental monitoring platform server, providing accurate data support and decision-making basis for environmental monitoring and governance; finally, through real-time data updates and analysis, the radar data analysis system can also provide dynamic and real-time monitoring and early warning services for environmental monitoring and governance.

[0029] In summary, the application of radar data analysis system in environmental monitoring and drone radar data analysis has significant advantages and value. It not only improves the accuracy and timeliness of environmental monitoring, but also provides strong technical support for environmental protection and sustainable development. With the continuous advancement of technology and the continuous expansion of application scenarios, it is believed that radar data analysis system will play an important role in more fields and promote the sustainable development of environmental monitoring.

[0030] In combination with the above application scenarios, each step in step 110 to step 140 is described in detail below.

[0031] When it comes to the technical solution recorded in step 110, several core concepts need to be clarified first: target spatiotemporal knowledge relationship space, target knowledge vector entity, target association vector entity, target linkage entity and target area element entity. These concepts together constitute the cornerstone of the radar data analysis system for processing airborne radar three-dimensional scanning data.

[0032] The target spatiotemporal knowledge relationship space is a multi-dimensional conceptual framework that integrates the three dimensions of space, time and knowledge to describe and store various information related to a specific target. In this space, each target (such as an airborne radar scan) is assigned a unique spatiotemporal coordinate, and various knowledge related to it (such as scan data characteristics, surface environment information, etc.) is also systematically organized and associated by radar data analysis.

[0033] The target knowledge vector entity is a basic unit in the target spatiotemporal knowledge relationship space, which represents the abstract representation of the airborne radar 3D scanning data at a specific time and location. This entity not only contains the direct information of the scanning data (such as distance, angle, intensity, etc.), but also implies the specific meaning and value of this data in the spatiotemporal context.

[0034] When the radar data analysis system executes step 110, it first goes deep into the target spatiotemporal knowledge relationship space to find and lock the target knowledge vector entity corresponding to the airborne radar 3D scanning data to be processed. This process is similar to accurately locating a bright pearl in the vast ocean of information.

[0035] Next, the radar data analysis system begins to explore the target association vector entities that are closely related to this target knowledge vector entity. These association vector entities are bridges connecting different data fragments and knowledge nodes, and they reveal the inherent connections and mutual influences between data. In this step, the target association vector entities are mainly divided into two categories:

[0036] (1) Target linkage entities: These entities represent other radar scanning data that have a spatiotemporal linkage relationship with the airborne radar 3D scanning data to be processed. They may be scanning data acquired at different time points or different spatial locations, but together reflect the environmental changes of the same target or the same area. By analyzing these target linkage entities, the radar data analysis system can capture the dynamic evolution of the surface environment, such as the spread of water pollution and changes in vegetation growth.

[0037] (2) Target area element entities: These entities directly correspond to the surface area elements covered by the airborne radar 3D scanning data to be processed, such as water bodies, vegetation, buildings, etc. Each target area element entity carries rich geographic information and characteristic attributes, such as the depth, flow rate, and water quality of water bodies; the type, distribution, and health of vegetation; the structure, height, and purpose of buildings, etc. By analyzing these target area element entities, the radar data analysis system can deeply understand the composition and characteristics of the surface environment.

[0038] During the execution of step 110, the radar data analysis system successfully extracts the target knowledge vector entity corresponding to the airborne radar 3D scanning data to be processed and its associated target linkage entity and target area element entity from the target spatiotemporal knowledge relationship space by comprehensively applying advanced data mining and association analysis techniques. This process not only lays a solid foundation for subsequent semantic mining and integrated processing, but also provides comprehensive and in-depth data support for environmental monitoring and governance. Through this series of complex and sophisticated operations, the radar data analysis system demonstrates its outstanding ability and great potential in processing massive and complex radar scanning data.

[0039] Further, step 120 is another core link in the radar data analysis system, which involves in-depth multi-modal semantic mining of target knowledge vector entities and target association vector entities. The purpose of this step is to extract rich, multi-level semantic information from these data entities to provide strong data support for subsequent environmental monitoring and governance.

[0040] Multimodal semantic mining is an advanced data analysis technology that can perform comprehensive semantic understanding and mining on different types of data. In radar data analysis systems, multimodal semantic mining is applied to process airborne radar 3D scanning data. By analyzing the implicit information in these data, radar data analysis systems can reveal the complex characteristics and dynamic changes of the surface environment.

[0041] When executing step 120, the radar data analysis system first mines the target knowledge vector entity to obtain its corresponding original multi-modal entity semantics. These semantic information covers the direct features of the scan data, such as distance, angle, intensity, etc., and also includes the deeper meaning of these data in the spatiotemporal context, such as surface morphology, target type, etc. By extracting the original multi-modal entity semantics, the radar data analysis system can provide rich basic data for subsequent analysis and processing.

[0042] Secondly, the radar data analysis system mines the target linkage entities and obtains the target multi-mode linkage semantics. These semantic information reflects the intrinsic connection and mutual influence between radar scanning data at different time points or spatial locations. By analyzing the target multi-mode linkage semantics, the radar data analysis system can capture the dynamic evolution of the surface environment, such as the path of water pollution diffusion and the trend of vegetation growth changes. This information is of great significance for environmental monitoring and governance, and can help relevant departments better understand the development trend and impact range of environmental problems.

[0043] Finally, the radar data analysis system also mines the target area element entities to obtain the target area element semantics. These semantic information describe in detail the characteristics and attributes of the surface area elements, such as the depth, flow rate, and water quality of water bodies; the type, distribution, and health status of vegetation, etc. By extracting the semantics of the target area elements, the radar data analysis system can provide comprehensive and accurate surface environmental information for environmental monitoring and governance, and support relevant departments to make more scientific and reasonable decisions.

[0044] It can be seen that step 120 successfully extracts rich semantic information from the target knowledge vector entity and the target association vector entity through multi-modal semantic mining technology. This information not only includes the direct features of the radar scanning data, but also covers the deep-level features and dynamic changes of the surface environment. Through the execution of this step, the radar data analysis system provides comprehensive and in-depth data support for environmental monitoring and governance, helping relevant departments to better understand and respond to environmental problems.

[0045] Next, step 130 is a key link in the radar data analysis system, which involves the integrated processing of the original multi-mode entity semantics, the target multi-mode linkage semantics and the target area element semantics to obtain the target multi-mode entity semantics of the airborne radar 3D scanning data to be processed. The purpose of this step is to integrate the multi-level and multi-dimensional semantic information previously mined to form a comprehensive and in-depth data representation, providing strong support for subsequent environmental monitoring and governance.

[0046] First, the radar data analysis system collects and organizes the original multi-modal entity semantics mined in the previous steps. These semantic information covers the direct characteristics of the airborne radar 3D scanning data and the deeper meaning in the spatiotemporal context, providing rich basic data for integrated processing.

[0047] Secondly, the radar data analysis system will take into account the multi-mode linkage semantics of the target. These semantic information reflects the intrinsic connection and mutual influence between radar scanning data at different time points or spatial locations, providing important clues for understanding the dynamic evolution of the surface environment. By combining these linkage semantics with the original entity semantics, the radar data analysis system can more comprehensively reveal the characteristics and changes of the scan data to be processed in the time and space dimensions.

[0048] Finally, the radar data analysis system also incorporates the semantics of target area elements into the integrated processing process. These semantic information describes in detail the characteristics and attributes of surface area elements, providing comprehensive and accurate surface environmental information for environmental monitoring and governance. By combining regional element semantics with other semantic information, the radar data analysis system can generate a comprehensive data representation that contains rich surface environmental characteristics and dynamic changes.

[0049] After this series of integrated processing steps, the radar data analysis system finally obtains the target multi-modal entity semantics of the airborne radar 3D scanning data to be processed. This comprehensive data representation not only includes the direct features and spatiotemporal background information of the scanning data, but also covers the deep-level features and dynamic change process of the surface environment. Through the execution of this step, the radar data analysis system provides comprehensive and in-depth data support for environmental monitoring and governance, helping relevant departments to better understand and respond to environmental problems.

[0050] Finally, step 140 is another key link in the radar data analysis system, which involves interaction with the environmental monitoring platform server to determine the degree of match between the airborne radar three-dimensional scanning data to be processed and the target environmental monitoring task, and sends the data with high matching degree to the corresponding server for further processing.

[0051] First, the radar data analysis system obtains the multimodal entity semantics corresponding to the environmental monitoring platform server in the target spatiotemporal knowledge relationship space. These multimodal entity semantics represent the environmental characteristics, spatiotemporal scope, and monitoring tasks that the environmental monitoring platform server is concerned about. By obtaining this information, the radar data analysis system can understand the current monitoring needs and priorities of the server.

[0052] Secondly, the radar data analysis system will calculate the matching degree based on the target multi-modal entity semantics obtained in the previous step and the multi-modal entity semantics corresponding to the environmental monitoring platform server. This calculation process will comprehensively consider the characteristics of the scan data, the spatiotemporal background, and the relevance to the environmental monitoring task to determine whether the airborne radar 3D scan data to be processed meets the monitoring requirements of the server.

[0053] During this process, the radar data analysis system will calculate a target matching probability, which reflects the matching degree between the scan data and the monitoring task of the environment monitoring platform server. The higher the target matching probability, the more the scan data meets the monitoring requirements of the server.

[0054] Finally, the radar data analysis system will send the pending airborne radar 3D scanning data with a target matching probability greater than the threshold to the corresponding environmental monitoring platform server according to the set probability threshold. In this way, the server can receive radar scanning data that is highly relevant to its monitoring task for further analysis and processing.

[0055] Through the execution of step 140, the radar data analysis system can achieve effective interaction with the environmental monitoring platform server, and timely send high-quality radar scanning data to the required server, providing strong data support for environmental monitoring and governance. At the same time, this step also reflects the excellent ability and huge potential of the radar data analysis system in processing massive and complex radar scanning data.

[0056] By applying the embodiments of the present application, a series of innovative steps are taken to greatly enhance the application value of airborne radar three-dimensional scanning data in environmental monitoring and governance. Specifically, the embodiments of the present application can first comprehensively and deeply mine the multi-level and multi-dimensional semantic information of radar scanning data in the target spatiotemporal knowledge relationship space, including original multi-modal entity semantics, target multi-modal linkage semantics, and target area element semantics. In this process, not only the direct characteristics of the scanning data are considered, but also its spatiotemporal background and its association with environmental elements are covered, thereby forming a comprehensive and dynamic understanding of the surface environment.

[0057] Furthermore, the embodiment of the present application integrates and processes these rich semantic information to obtain the target multi-mode entity semantics of the airborne radar three-dimensional scanning data to be processed. This comprehensive data representation significantly enhances the ability of the scanning data to describe and interpret the surface environment. In addition, the embodiment of the present application also realizes effective interaction with the environmental monitoring platform server. By calculating the target matching probability, high-quality radar scanning data is sent to the required server in a timely manner, providing more accurate and timely data support for environmental monitoring and governance.

[0058] In general, the embodiments of the present application solve the problems existing in traditional radar data analysis, such as insufficient semantic information mining and low matching degree between data and environmental monitoring tasks, and significantly improve the application effect and value of radar scanning data in environmental monitoring and governance.

[0059] In some optional embodiments, the method is implemented by a target radar data matching analysis algorithm, and the debugging step of the target radar data matching analysis algorithm includes: obtaining a spatiotemporal knowledge relationship space generated according to an environmental monitoring log report in an environmental monitoring database, the spatiotemporal knowledge relationship space including a plurality of mapping entities, each of the mapping entities corresponding to the environmental monitoring log report, and the environmental monitoring log report including historical airborne radar three-dimensional scanning data and surface area elements of the historical airborne radar three-dimensional scanning data; from the spatiotemporal knowledge relationship space, determining a knowledge vector entity corresponding to the target historical airborne radar three-dimensional scanning data in the historical airborne radar three-dimensional scanning data and an associated vector entity of the knowledge vector entity, the associated vector entity including a linkage radar scanning data corresponding to the linkage radar scanning data involved in the target historical airborne radar three-dimensional scanning data. The invention discloses a method for mining the surface area elements of the target historical airborne radar three-dimensional scanning data and the regional element entity corresponding to the surface area elements of the target historical airborne radar three-dimensional scanning data; multi-mode semantic mining is performed on the knowledge vector entity and the association vector entity respectively to obtain the airborne radar scanning data semantics corresponding to the knowledge vector entity, the multi-mode linkage semantics corresponding to the linkage entity and the surface area element semantics corresponding to the regional element entity; the airborne radar scanning data semantics, the multi-mode linkage semantics and the regional element semantics are integrated to obtain the multi-mode entity semantic sample of the target historical airborne radar three-dimensional scanning data; based on the multi-mode entity semantic sample, the algorithm weights in the initial radar data matching analysis algorithm are optimized to obtain the pre-debugging radar data matching analysis algorithm; based on the multi-mode entity semantic sample, the pre-debugging radar data matching analysis algorithm is debugged to obtain the target radar data matching analysis algorithm.

[0060] It can be understood that the method involved in the embodiment of the present application is implemented by a target radar data matching and analysis algorithm. The core of the algorithm is to ensure that the algorithm can accurately match and analyze the airborne radar three-dimensional scanning data through a series of rigorous debugging steps, thereby providing strong support for environmental monitoring and governance.

[0061] First, the radar data analysis system needs to obtain a large number of environmental monitoring log reports from the environmental monitoring database. These reports are the basis for constructing the spatiotemporal knowledge relationship space. The spatiotemporal knowledge relationship space is a complex structure containing several mapping entities, each of which corresponds to an environmental monitoring log report. These log reports not only contain historical airborne radar 3D scanning data, but also describe in detail the surface area elements corresponding to these data, such as terrain, vegetation, water bodies, etc. This design enables the algorithm to fully consider the relationship between radar scanning data and its spatiotemporal background, providing a rich information foundation for subsequent multimodal semantic mining.

[0062] Next, the radar data analysis system needs to determine the knowledge vector entities and association vector entities corresponding to the target's historical airborne radar 3D scanning data from the spatiotemporal knowledge relationship space. The knowledge vector entity is a direct representation of the radar scanning data in the spatiotemporal knowledge relationship space, while the association vector entity further describes other information related to the radar scanning data, including the linkage entity corresponding to the radar scanning data with which it has a linkage relationship, and the regional element entity corresponding to the surface regional element corresponding to the radar scanning data. By extracting these knowledge vector entities and association vector entities, the radar data analysis system can more comprehensively understand the characteristics and changes of the radar scanning data in the spatiotemporal context.

[0063] Then, the radar data analysis system needs to perform multi-modal semantic mining on the extracted knowledge vector entities and associated vector entities. This is a process of deeply analyzing the intrinsic meaning of the data. Through mining, the radar data analysis system can obtain the semantics of the airborne radar scanning data corresponding to the knowledge vector entity, which is a direct interpretation of the radar scanning data in the spatiotemporal context; at the same time, the radar data analysis system can also obtain the multi-modal linkage semantics corresponding to the linkage entity, which reveals the intrinsic connection and mutual influence between different radar scanning data; finally, the radar data analysis system can also obtain the surface regional element semantics corresponding to the regional element entity, which describes in detail the characteristics and attributes of the surface regional elements. Through the execution of this step, the radar data analysis system can fully reveal the intrinsic meaning and spatiotemporal context of the radar scanning data from multiple dimensions and levels.

[0064] After obtaining these rich semantic information, the radar data analysis system needs to integrate them to form a comprehensive and in-depth understanding of the historical airborne radar 3D scanning data of the target. The process of integrated processing is to combine the semantics of airborne radar scanning data, multi-mode linkage semantics and regional element semantics to form a comprehensive data representation containing multi-level and multi-dimensional information. This comprehensive data representation is called a multi-mode entity semantic sample, which fully reflects the characteristics and change process of radar scanning data in the spatiotemporal context.

[0065] With the multimodal entity semantic samples, the radar data analysis system can start to utilize and optimize them. First, the algorithm will optimize the algorithm weights in the initial radar data matching analysis algorithm based on the multimodal entity semantic samples. This is an iterative adjustment process, by constantly adjusting the algorithm weights, so that the algorithm can better match and analyze radar scan data. After the optimization of this step, the performance of the algorithm will be significantly improved, forming a pre-debug radar data matching analysis algorithm.

[0066] Finally, the radar data analysis system also needs to further debug and improve the pre-debugging radar data matching analysis algorithm. The purpose of this step is to ensure that the algorithm can achieve the expected results in practical applications. Through the debugging process based on multi-modal entity semantic samples, the algorithm can continuously correct its own deviations and deficiencies, and finally form a stable and reliable target radar data matching analysis algorithm. This algorithm can accurately match and analyze airborne radar 3D scanning data, providing strong data support for environmental monitoring and governance.

[0067] In this design, we first generate a spatiotemporal knowledge relationship space by obtaining environmental monitoring log reports; then extract knowledge vector entities and associated vector entities corresponding to the target historical airborne radar 3D scanning data from the space; then perform multimodal semantic mining on these entities to obtain rich semantic information; then integrate these semantic information to form multimodal entity semantic samples; finally, optimize and debug the initial algorithm based on the multimodal entity semantic samples to obtain the target algorithm. This series of steps ensures that the algorithm can accurately match and analyze airborne radar 3D scanning data to provide strong support for environmental monitoring and governance.

[0068] In the following embodiment, determining the knowledge vector entity corresponding to the target historical airborne radar three-dimensional scanning data in the historical airborne radar three-dimensional scanning data and the associated vector entity of the knowledge vector entity from the spatiotemporal knowledge relationship space includes: obtaining a plurality of knowledge transfer strategies, the knowledge transfer strategies including different entity index tags; based on the different entity index tags, determining the knowledge vector entity corresponding to the target historical airborne radar three-dimensional scanning data in the historical airborne radar three-dimensional scanning data and the associated vector entity of the knowledge vector entity from the spatiotemporal knowledge relationship space.

[0069] It can be understood that this embodiment involves how to determine the knowledge vector entity and its associated vector entity corresponding to the target data in the historical airborne radar three-dimensional scanning data from the spatiotemporal knowledge relationship space. The core of this process is to use a series of carefully designed knowledge transmission strategies to guide the radar data analysis system to find the accurate target in the complex spatiotemporal knowledge relationship space.

[0070] First, the radar data analysis system needs to obtain several knowledge transfer strategies. These strategies are bridges connecting various entities in the spatiotemporal knowledge relationship space. They contain different entity index tags, which are like identity cards, uniquely and accurately identifying each entity in the spatiotemporal knowledge relationship space. The formulation of knowledge transfer strategies is based on a deep understanding of historical airborne radar 3D scanning data and a precise grasp of the relationship between these data and the spatiotemporal background. Through these strategies, the radar data analysis system can effectively navigate in the spatiotemporal knowledge relationship space and find the knowledge vector entities and their associated vector entities required by the radar data analysis system.

[0071] With the knowledge transfer strategy, the radar data analysis system can begin to determine the knowledge vector entity corresponding to the target's historical airborne radar 3D scanning data from the spatiotemporal knowledge relationship space. This process is like looking for a specific location in a complex map. The radar data analysis system needs to use the navigation tool in hand (i.e., the knowledge transfer strategy) to guide the radar data analysis system forward. In this process, the radar data analysis system will compare the entities in the spatiotemporal knowledge relationship space one by one according to the entity index label in the knowledge transfer strategy until it finds the knowledge vector entity that completely matches the target's historical airborne radar 3D scanning data.

[0072] Finding the knowledge vector entity is only the first step. The radar data analysis system needs to further determine the associated vector entity of this entity. The associated vector entity is an entity that has some connection or relationship with the knowledge vector entity. Together, they constitute a network in the spatiotemporal knowledge relationship space. The process of determining the associated vector entity is like exploring the environment and traffic routes around the destination after finding it. The radar data analysis system will analyze the association relationship between the knowledge vector entity and other entities according to the guidance in the knowledge transfer strategy, so as to find all the associated vector entities related to the knowledge vector entity.

[0073] In this process, the knowledge transmission strategy plays a vital role. It not only provides accurate navigation direction for the radar data analysis system, but also helps the radar data analysis system avoid getting lost and wandering in the spatiotemporal knowledge relationship space. By following these strategies, the radar data analysis system can efficiently find the knowledge vector entities and their associated vector entities corresponding to the target historical airborne radar 3D scanning data, providing a solid foundation for subsequent multi-modal semantic mining and integrated processing.

[0074] In addition, the flexibility and diversity of the knowledge transfer strategy is also a major advantage. In practical applications, the radar data analysis system can adjust the content and form of the strategy according to specific needs and scenarios to adapt to different spatiotemporal knowledge relationship spaces and different historical airborne radar 3D scanning data. This flexibility and diversity makes the knowledge transfer strategy a powerful and versatile tool, providing unlimited possibilities for the radar data analysis system to explore and navigate in complex spatiotemporal knowledge relationship spaces.

[0075] It can be seen that determining the knowledge vector entities and their associated vector entities corresponding to the target data in the historical airborne radar 3D scanning data from the spatiotemporal knowledge relationship space is a complex and important process. By acquiring and utilizing several carefully designed knowledge transfer strategies, the radar data analysis system can effectively navigate and explore in the spatiotemporal knowledge relationship space to find the accurate target entity and its associated entities. This process not only provides a solid foundation for subsequent multimodal semantic mining and integrated processing, but also demonstrates the great potential and application value of knowledge transfer strategies in complex data processing and analysis.

[0076] In some examples, the different entity index tags include a head entity index tag and a preceding and following entity index tags; based on the different entity index tags, determining the knowledge vector entity corresponding to the target historical airborne radar three-dimensional scanning data in the historical airborne radar three-dimensional scanning data and the associated vector entity of the knowledge vector entity from the spatiotemporal knowledge relationship space, including: determining the mapping entity corresponding to the head entity index tag from the spatiotemporal knowledge relationship space, and obtaining the knowledge vector entity corresponding to the target historical airborne radar three-dimensional scanning data; obtaining the original associated vector entity corresponding to the knowledge vector entity; determining the original associated vector entity corresponding to the preceding and following entity index tags from the original associated vector entity, and obtaining the associated vector entity of the knowledge vector entity.

[0077] Before referring to the technical solution described in this embodiment, it is necessary to first understand several core concepts: spatiotemporal knowledge relationship space, head entity index label, preceding and following entity index labels, knowledge vector entity and associated vector entity. The spatiotemporal knowledge relationship space is a data structure containing a large number of entities and their relationships, in which each entity has a unique index label, which is used to uniquely identify the entity in the space. The head entity index label specifically refers to the entity index label that refers to the subject or starting point in a certain relationship, while the preceding and following entity index labels refer to the index labels of other entities that have an order or association with the head entity in a certain specific relationship. The knowledge vector entity is a specific entity in the spatiotemporal knowledge relationship space, which is represented in the form of a vector and contains rich semantic information. The associated vector entity refers to other vector entities that have some kind of association or relationship with the knowledge vector entity.

[0078] Based on the above concepts, the technical solution of this embodiment aims to determine the knowledge vector entity corresponding to the target historical airborne radar 3D scanning data from the spatiotemporal knowledge relationship space, and further find the associated vector entity of the knowledge vector entity. The realization of this process not only depends on the in-depth understanding of the spatiotemporal knowledge relationship space, but also requires the precise grasp of the role of different entity index tags in the relationship definition.

[0079] First, from the spatiotemporal knowledge relationship space, the mapping entity corresponding to the head entity index label of the target historical airborne radar 3D scanning data is determined. This step is the key starting point, which requires the system to accurately identify the specific entity pointed to by the head entity index label, that is, the knowledge vector entity corresponding to the target historical airborne radar 3D scanning data. The head entity index label plays a bridge role here, connecting the external input (that is, the target historical airborne radar 3D scanning data) and the entity representation inside the spatiotemporal knowledge relationship space. Through this mapping process, the system can convert external specific data into internally operable vector entities, laying the foundation for subsequent relational reasoning and entity association analysis.

[0080] Secondly, the original association vector entity corresponding to the knowledge vector entity is obtained. This step is the process of further analyzing the determined knowledge vector entity, with the aim of finding other entities that are directly associated with the entity. In the spatiotemporal knowledge relationship space, the associations between entities are complex and diverse, and may include temporal precedence, spatial proximity, and attribute similarity. Therefore, the implementation of this step requires the system to have a strong relational reasoning capability, which can filter out the most direct and relevant ones from many possible associations to the target knowledge vector entity.

[0081] Finally, the original associated vector entities corresponding to the preceding and following entity index labels are determined from the original associated vector entities, thereby obtaining the associated vector entities of the knowledge vector entity. This step is the process of screening and refining the original associated vector entities obtained in the previous step. The preceding and following entity index labels act as a filter here, which helps the system select those entities that meet specific order or association conditions from a large number of original associated vector entities as the final associated vector entities of the knowledge vector entity. The implementation of this step requires not only the system to have an accurate understanding and application of the preceding and following entity index labels, but also the system to be able to accurately track and locate qualified entities in a complex association network.

[0082] In this way, the knowledge vector entity corresponding to the target historical airborne radar 3D scanning data is determined from the spatiotemporal knowledge relationship space, and the associated vector entity of the knowledge vector entity is further found. The realization of this process not only demonstrates the powerful ability of the system in processing complex spatiotemporal data, but also provides a solid foundation for subsequent data analysis, pattern recognition, decision support and other applications.

[0083] In terms of specific technical implementation, the solution may involve a variety of advanced artificial intelligence and machine learning technologies, such as deep learning, natural language processing, and graph neural networks. Through the comprehensive application of these technologies, the system can efficiently process and analyze large-scale spatiotemporal knowledge relationship data, extract valuable information and knowledge from it, and provide strong support for various practical application scenarios. For example, in the processing and analysis of historical airborne radar 3D scanning data, the solution can help the system better understand and identify the spatiotemporal relationships and association patterns between different scanning data, thereby providing a scientific basis for decision-making in the fields of environmental monitoring and disaster warning.

[0084] Under some preferred design ideas, the associated vector entity includes multiple layers of associated vector entities, the linkage entity and the regional element entity are the first layer of associated vector entities of the knowledge vector entity; the airborne radar scanning data semantics, the multi-mode linkage semantics and the regional element semantics are integrated to obtain a multi-mode entity semantic sample of the target historical airborne radar three-dimensional scanning data, including: determining the bottom-level associated vector entity of the knowledge vector entity from the multiple layers of associated vector entities, and determining the upper-level associated vector entity of the bottom-level associated vector entity from the multiple layers of associated vector entities; integrating the multi-mode entity semantics of the bottom-level associated vector entity and the multi-mode entity semantics of the upper-level associated vector entity to obtain a multi-mode entity semantic sample of the upper-level associated vector entity. Example; if the upper layer association vector entity is not the first layer association vector entity, then the multimodal entity semantic sample of the upper layer association vector entity is used as the multimodal entity semantics of the upper layer association vector entity, and the upper layer association vector entity is used as the bottom layer association vector entity, and the process jumps to the step of determining the upper layer association vector entity of the bottom layer association vector entity from the multiple layers of association vector entities; if the upper layer association vector entity is the first layer association vector entity, then the multimodal entity semantic sample of the upper layer association vector entity is used as the multimodal linkage semantics or the regional element semantics; the airborne radar scanning data semantics, the multimodal linkage semantics and the regional element semantics are integrated and processed to obtain the multimodal entity semantic sample of the target historical airborne radar three-dimensional scanning data.

[0085] Under this design concept, the association vector entity is further refined into a multi-layer association vector entity, which means that there is a hierarchical structure between the association vector entities, and each layer represents a relationship with different depths and breadths with the knowledge vector entity. Furthermore, the goal of this embodiment is to determine the knowledge vector entity corresponding to the target data in the historical airborne radar three-dimensional scanning data, and further obtain the multi-layer association vector entity of the knowledge vector entity, and finally obtain the multi-mode entity semantic sample of the target historical airborne radar three-dimensional scanning data through the integrated processing of these entity semantics.

[0086] First, the knowledge vector entity corresponding to the target's historical airborne radar 3D scanning data is determined from the spatiotemporal knowledge relationship space. This process is completed through matching and mapping. The system will find the unique knowledge vector entity corresponding to it in the spatiotemporal knowledge relationship space based on the characteristics of the target data.

[0087] Secondly, we start to process the multi-layer associative vector entities of the knowledge vector entities. The concept of multi-layer associative vector entities plays a key role here, which allows the system to understand and process the relationship between the knowledge vector entity and other entities in a hierarchical manner. For example, the system will determine the bottom-level associative vector entities of the knowledge vector entity from the multi-layer associative vector entities. These bottom-level associative vector entities represent the most direct and basic relationship with the knowledge vector entity. Then, the system will further determine the upper-level associative vector entities of these bottom-level associative vector entities. These upper-level associative vector entities represent slightly more indirect relationships with the knowledge vector entity, but still belong to the important category of association.

[0088] Next, the multimodal entity semantics of the underlying association vector entity and the multimodal entity semantics of the upper-layer association vector entity are integrated. Multimodal entity semantics refers to the semantic representation of an entity in different modes or dimensions. Integration processing is to fuse the semantic information in these different modes or dimensions to obtain a more comprehensive and richer semantic representation. This process is completed through complex semantic calculation and analysis. The system will effectively integrate the multimodal entity semantics of the underlying and upper-layer association vector entities based on the relationship and semantic similarity between the entities.

[0089] Then, it is checked whether the previous layer association vector entity is a first layer association vector entity. First layer association vector entities have a special status here, and they represent the most direct and important association with the knowledge vector entity. If the previous layer association vector entity is not a first layer association vector entity, the system will take its multimodal entity semantic sample as its new multimodal entity semantics and take it as the new bottom layer association vector entity, and then jump back to the step of determining the previous layer association vector entity of the bottom layer association vector entity from the multi-layer association vector entity. This process will be iterated until the previous layer association vector entity becomes a first layer association vector entity.

[0090] Finally, when the upper layer of associated vector entities is the first layer of associated vector entities, its multimodal entity semantic sample will be used as multimodal linkage semantics or regional element semantics. Multimodal linkage semantics and regional element semantics are two important concepts in this scheme, which represent semantic information closely related to the historical airborne radar 3D scanning data of the target. Then, the system will perform final integration processing on the airborne radar scanning data semantics, multimodal linkage semantics and regional element semantics to obtain the multimodal entity semantic sample of the historical airborne radar 3D scanning data of the target. This process is the core goal of the scheme. It obtains a comprehensive, rich and accurate semantic representation of the target data by effectively integrating and fusing semantic information from different sources and modes.

[0091] In this way, the knowledge vector entity corresponding to the target historical airborne radar 3D scanning data is determined from the spatiotemporal knowledge relationship space, and the multi-layer associated vector entity of the knowledge vector entity is further obtained and processed. Finally, through the integrated processing of these entity semantics, a multi-modal entity semantic sample of the target data is obtained. The realization of this process not only demonstrates the system's powerful ability to process complex spatiotemporal data, but also provides a solid foundation for subsequent data analysis, pattern recognition, decision support and other applications.

[0092] In an alternative embodiment, the determining of the underlying association vector entity of the knowledge vector entity from the multi-layer association vector entity, and the determining of the upper-layer association vector entity of the underlying association vector entity from the multi-layer association vector entity, includes: acquiring a knowledge transfer strategy, the knowledge transfer strategy including a preceding and following entity index tag; based on the preceding and following entity index tags, determining the underlying association vector entity of the knowledge vector entity from the multi-layer association vector entity, and determining the upper-layer association vector entity of the underlying association vector entity from the multi-layer association vector entity.

[0093] In this embodiment, the bottom-level associated vector entities refer to those vector entities that are most directly and fundamentally associated with the knowledge vector entity, and they are at the bottom level in the multi-layer associated vector entity. Further, the goal of this embodiment is to determine the knowledge vector entity corresponding to the target data in the historical airborne radar three-dimensional scanning data from the spatiotemporal knowledge relationship space, and further determine the bottom-level associated vector entity of the knowledge vector entity and the upper-level associated vector entity of the bottom-level associated vector entity.

[0094] First, we need to obtain the knowledge transfer strategy. The knowledge transfer strategy is a key concept that guides how to determine the bottom-level association vector entity and the upper-level association vector entity from multiple layers of association vector entities. This strategy contains the index labels of the preceding and following entities, which are important bases for determining the order and hierarchy of association vector entities.

[0095] After acquiring the knowledge transfer strategy, the embodiment begins to determine the underlying association vector entities of the knowledge vector entity from the multi-layer association vector entity based on the preceding and following entity index tags. This process is completed through matching and mapping. The system will search for underlying association vector entities that match the knowledge vector entity and conform to the preceding and following entity index tags in the multi-layer association vector entity based on the characteristics of the knowledge vector entity. These underlying association vector entities represent the most direct and basic relationship with the knowledge vector entity and are an important basis for further analysis and understanding of the knowledge vector entity.

[0096] After determining the bottom-level association vector entities, the embodiment further determines the upper-level association vector entities of these bottom-level association vector entities from the multi-level association vector entities. The association vector entities of this layer are slightly higher in level than the bottom-level association vector entities, but they still have important associations with the knowledge vector entities. The system can accurately determine these upper-level association vector entities based on the preceding and following entity index tags and the relationship between the bottom-level association vector entities and the upper-level association vector entities.

[0097] In this process, the preceding and following entity index labels play a crucial role. They not only guide how to determine the bottom-level associated vector entities from the multi-level associated vector entities, but also ensure that the order and hierarchical relationship between the determined bottom-level associated vector entities and the upper-level associated vector entities are correct. This enables the system to more accurately understand and analyze the relationship between the knowledge vector entity and its associated vector entity, providing a solid foundation for subsequent data analysis, pattern recognition, decision support and other applications.

[0098] In this way, by acquiring the knowledge transmission strategy and determining the bottom-level associated vector entity and the upper-level associated vector entity of the knowledge vector entity from the multi-layer associated vector entity based on the preceding and following entity index tags, the knowledge vector entity and its associated vector entity corresponding to the target data in the historical airborne radar 3D scanning data are accurately determined. The realization of this process not only demonstrates the system's powerful ability to process complex spatiotemporal data, but also provides more possibilities and options for subsequent data analysis, pattern recognition, decision support and other applications.

[0099] In some other replaceable embodiments, the multimodal entity semantics of the underlying association vector entity and the multimodal entity semantics of the upper-layer association vector entity are integrated to obtain a multimodal entity semantic sample of the upper-layer association vector entity, including: obtaining the correlation between the underlying association vector entity and the upper-layer association vector entity, the correlation representing the influence coefficient of the underlying association vector entity on the upper-layer association vector entity; updating the multimodal entity semantics of the underlying association vector entity based on the correlation to obtain updated multimodal entity semantics of the underlying association vector entity; and integrating the updated multimodal entity semantics with the multimodal entity semantics of the upper-layer association vector entity to obtain a multimodal entity semantic sample of the upper-layer association vector entity.

[0100] Furthermore, the obtaining of the correlation between the underlying association vector entity and the upper-layer association vector entity includes: determining a feature commonality value between the upper-layer association vector entity and the underlying association vector entity based on the multimodal entity semantics of the underlying association vector entity and the multimodal entity semantics of the upper-layer association vector entity; and determining the correlation between the underlying association vector entity and the upper-layer association vector entity based on the feature commonality value.

[0101] In this embodiment, the process of integrating the multimodal entity semantics of the underlying association vector entity and the multimodal entity semantics of the upper-layer association vector entity to obtain the multimodal entity semantic sample of the upper-layer association vector entity is given more careful consideration and strategy. This process is not just a simple merger of the two semantics, but a deep fusion based on the internal relationship between them.

[0102] First, the radar data analysis system focuses on obtaining the correlation between the underlying association vector entity and the upper-layer association vector entity. This correlation is not given arbitrarily, but has a clear physical or logical meaning, that is, it represents the influence coefficient of the underlying association vector entity on the upper-layer association vector entity. This coefficient is the key to understanding the relationship between the two, and it determines how much weight or strength the multimodal entity semantics of the underlying association vector entity should be considered during integration processing.

[0103] To get this correlation, the radar data analysis system takes a series of delicate steps. It first determines the feature commonality value between the bottom-level association vector entity and the top-level association vector entity based on the multimodal entity semantics of the two. This feature commonality value is a quantitative indicator that reflects the similarity or number of shared features between the two entities at the semantic level. In other words, it measures the degree of "resonance" between the bottom-level association vector entity and the top-level association vector entity in terms of multimodal entity semantics.

[0104] After determining the feature commonality value, the radar data analysis system further determines the correlation between the underlying association vector entity and the upper-layer association vector entity based on this value. This step is to convert the feature commonality value into an actual influence coefficient, which will play a key role in the subsequent integration processing. Through this step, the radar data analysis system is actually building a precise "recipe" for integration processing to ensure that the multimodal entity semantics of the underlying association vector entity can be integrated into the multimodal entity semantics of the upper-layer association vector entity in an appropriate manner and degree.

[0105] Next, the radar data analysis system updates the multimodal entity semantics of the underlying association vector entity based on the previously obtained correlation. This update process is not a simple modification or adjustment, but a purposeful enhancement or weakening of the multimodal entity semantics of the underlying association vector entity based on the correlation. The purpose of this is to ensure that during the integrated processing, the multimodal entity semantics of the underlying association vector entity can coexist more harmoniously with the multimodal entity semantics of the upper-layer association vector entity, thereby producing a richer, more comprehensive and accurate multimodal entity semantic sample of the upper-layer association vector entity.

[0106] Finally, the radar data analysis system integrates the updated multimodal entity semantics with the multimodal entity semantics of the upper-layer associated vector entity. This step is the climax of the entire process and the ultimate goal of all previous steps. Through integration processing, the radar data analysis system merges the updated multimodal entity semantics of the underlying associated vector entity with the multimodal entity semantics of the upper-layer associated vector entity to form a new multimodal entity semantic sample of the upper-layer associated vector entity that contains more information and details. This sample not only retains the main features of the original two semantics, but also introduces new and richer semantic elements through the correlation between them.

[0107] It can be seen that through a series of carefully designed steps, the multimodal entity semantics of the underlying associated vector entity and the multimodal entity semantics of the upper layer associated vector entity are deeply integrated. This process not only takes into account the intrinsic relationship between the two, but also ensures that the results of the integrated processing are both accurate and comprehensive through precise calculations and strategic adjustments. The implementation of this technical solution undoubtedly provides a new and more effective method and idea for entity semantic analysis in spatiotemporal knowledge relationship space.

[0108] In terms of specific technical implementation, the solution may involve a variety of advanced artificial intelligence and machine learning technologies. For example, it may require the use of deep learning technology to extract and represent the multimodal entity semantics of entities; natural language processing technology to understand and analyze the commonalities and differences between semantics; and graph neural networks and other technologies to model and reason about the relationships and influences between entities. Through the comprehensive application of these technologies, the radar data analysis system can efficiently process and analyze large-scale spatiotemporal knowledge relationship data, extract valuable information and knowledge from it, and provide strong support for various practical application scenarios.

[0109] In addition, by adjusting the calculation method of correlation and the strategy of integrated processing, it can easily adapt to different application scenarios and needs. For example, in different application scenarios, the correlation between the underlying correlation vector entity and the upper layer correlation vector entity may have different calculation methods and interpretations. In some scenarios, the correlation may be more based on the physical distance or time order between entities; while in other scenarios, it may be more based on the logical association or semantic similarity between entities. By flexibly adjusting these parameters and strategies, the radar data analysis system can easily cope with various complex application scenarios and provide users with more accurate and useful information and knowledge.

[0110] Thus, through a series of meticulous steps and strategic considerations, the deep integration of multimodal entity semantics of underlying association vector entities and multimodal entity semantics of upper-layer association vector entities is achieved. The realization of this process not only demonstrates the powerful ability in processing complex spatiotemporal data, but also provides more possibilities and options for subsequent data analysis, pattern recognition, decision support and other applications.

[0111] In an optional embodiment, the airborne radar scanning data semantics, the multi-mode linkage semantics and the regional element semantics are integrated to obtain a multi-mode entity semantic sample of the target historical airborne radar three-dimensional scanning data, including: integrating the airborne radar scanning data semantics and the multi-mode linkage semantics to obtain a first multi-mode entity semantic sample of the target historical airborne radar three-dimensional scanning data; integrating the airborne radar scanning data semantics and the regional element semantics to obtain a second multi-mode entity semantic sample of the target historical airborne radar three-dimensional scanning data; and determining a multi-mode entity semantic sample of the target historical airborne radar three-dimensional scanning data based on the first multi-mode entity semantic sample and the second multi-mode entity semantic sample.

[0112] Further, the determining of the multimodal entity semantic sample of the target historical airborne radar three-dimensional scanning data based on the first multimodal entity semantic sample and the second multimodal entity semantic sample includes: obtaining a first correlation corresponding to the first multimodal entity semantic sample and a second correlation corresponding to the second multimodal entity semantic sample; updating the first multimodal entity semantic sample based on the first correlation to obtain an updated first multimodal entity semantic sample; updating the second multimodal entity semantic sample based on the second correlation to obtain an updated second multimodal entity semantic sample; and determining the multimodal entity semantic sample of the target historical airborne radar three-dimensional scanning data based on the updated first multimodal entity semantic sample and the updated second multimodal entity semantic sample.

[0113] Furthermore, the method of obtaining the first correlation corresponding to the first multimodal entity semantic sample and the second correlation corresponding to the second multimodal entity semantic sample includes: performing a first integration process on the first multimodal entity semantic sample and the second multimodal entity semantic sample to obtain integrated multimodal entity semantics; performing a first attention enhancement on the integrated multimodal entity semantics to obtain first multimodal attention enhanced semantics; and determining the first correlation corresponding to the first multimodal entity semantic sample and the second correlation corresponding to the second multimodal entity semantic sample from the first multimodal attention enhanced semantics.

[0114] Based on this embodiment, the process of integrating and processing the semantics of airborne radar scanning data, multi-mode linkage semantics and regional element semantics to obtain multi-mode entity semantic samples of the target historical airborne radar three-dimensional scanning data is a multi-level and multi-dimensional information fusion and refinement process. This process not only requires the accurate capture and expression of the intrinsic characteristics of various semantic information, but also requires the ingenious processing of the complex relationships between them to ensure that the multi-mode entity semantic samples finally generated can fully and accurately reflect the rich connotations of the target historical airborne radar three-dimensional scanning data.

[0115] First, the radar data analysis system integrates the semantics of airborne radar scanning data and multi-mode linkage semantics to obtain the first multi-mode entity semantic sample of the target historical airborne radar 3D scanning data. This step is the basis of integrated processing. It combines the physical features such as space, shape, and texture contained in the airborne radar scanning data with the dynamic information such as time, events, and behaviors represented by the multi-mode linkage semantics, and initially constructs a semantic sample containing multi-dimensional information.

[0116] Next, the radar data analysis system integrates the semantics of the airborne radar scan data and the semantics of the regional elements to obtain the second multi-modal entity semantic sample of the target historical airborne radar 3D scan data. This step further enriches the connotation of the semantic sample by combining the airborne radar scan data with the semantics of the regional elements, integrating the local features reflected by the scan data with the overall environment, background, function and other information of the region, forming a more three-dimensional and comprehensive semantic sample.

[0117] Then, the radar data analysis system determines the multimodal entity semantic sample of the target historical airborne radar 3D scanning data based on the first multimodal entity semantic sample and the second multimodal entity semantic sample. This step is the core of the integrated processing, which requires the relationship between the two multimodal entity semantic samples to be handled ingeniously to ensure that they can coexist harmoniously in a unified semantic framework.

[0118] Furthermore, in order to determine the multimodal entity semantic examples of the target historical airborne radar three-dimensional scanning data, the radar data analysis system first obtains the first correlation corresponding to the first multimodal entity semantic example and the second correlation corresponding to the second multimodal entity semantic example. The two correlations represent the importance and contribution of the two multimodal entity semantic examples in the overall semantic framework.

[0119] Based on the first correlation, the radar data analysis system updates the first multimodal entity semantic sample to obtain an updated first multimodal entity semantic sample. This step is a further optimization and adjustment of the first multimodal entity semantic sample to ensure that it can better play its role in the overall semantic framework. Similarly, based on the second correlation, the radar data analysis system also updates the second multimodal entity semantic sample to obtain an updated second multimodal entity semantic sample.

[0120] Finally, the radar data analysis system determines the multimodal entity semantic sample of the target historical airborne radar 3D scanning data based on the updated first multimodal entity semantic sample and the updated second multimodal entity semantic sample. This step is the final result of the integrated processing, which combines the two optimized multimodal entity semantic samples to form a comprehensive, accurate and rich multimodal entity semantic sample.

[0121] Furthermore, in order to obtain the first correlation corresponding to the first multimodal entity semantic sample and the second correlation corresponding to the second multimodal entity semantic sample, the radar data analysis system takes a series of sophisticated steps. It first performs a first integration process on the first multimodal entity semantic sample and the second multimodal entity semantic sample to obtain the integrated multimodal entity semantics. This step is to initially combine the two multimodal entity semantic samples to form a semantic representation containing more information. Then, it performs a first attention enhancement on the integrated multimodal entity semantics to obtain the first multimodal attention enhanced semantics. This step is to make the overall semantic representation more prominent and significant by enhancing certain important semantic features. Finally, it determines the first correlation corresponding to the first multimodal entity semantic sample from the first multimodal attention enhanced semantics and the second correlation corresponding to the second multimodal entity semantic sample. This step is to extract the importance and contribution of the two multimodal entity semantic samples in the overall semantic framework by analyzing the enhanced semantic representation.

[0122] In this way, through multi-level and multi-dimensional information fusion and refinement, the integrated processing of airborne radar scanning data semantics, multi-mode linkage semantics and regional element semantics is realized, and a comprehensive, accurate and rich multi-mode entity semantic sample of the target historical airborne radar 3D scanning data is obtained. This process not only demonstrates the powerful ability in processing complex spatiotemporal data, but also provides more possibilities and options for subsequent data analysis, pattern recognition, decision support and other applications.

[0123] In some independent embodiments, the algorithm weights in the initial radar data matching analysis algorithm are optimized based on the multimodal entity semantic samples to obtain a pre-debugging radar data matching analysis algorithm, including: performing a second attention enhancement on the multimodal entity semantic samples to obtain a second multimodal attention enhancement semantics corresponding to the target historical airborne radar three-dimensional scanning data; performing a second integration processing on the algorithm weights and the second multimodal attention enhancement semantics to optimize the algorithm weights in the initial radar data matching analysis algorithm to obtain a pre-debugging radar data matching analysis algorithm.

[0124] It can be understood that the core of this embodiment is to optimize the algorithm weights in the initial radar data matching analysis algorithm using multi-modal entity semantic samples, thereby obtaining a pre-debugging radar data matching analysis algorithm. This process involves in-depth processing of multi-modal entity semantic samples and fine integration with algorithm weights, aiming to improve the accuracy and efficiency of the radar data matching analysis algorithm.

[0125] First of all, strengthening the second attention to multimodal entity semantic samples is an important step of this technical solution. The multimodal entity semantic samples contain rich information such as airborne radar scanning data, multimodal linkage semantics, and regional element semantics. However, the distribution of this information in the samples is not uniform. Some information is more important for radar data matching analysis, while some information is relatively less important. Therefore, through the second attention strengthening, those information that are more critical to radar data matching analysis can be highlighted, so that this information can play a greater role in the subsequent algorithm weight optimization. This step is similar to screening out the most important clues from a large amount of information, providing strong support for subsequent analysis and decision-making.

[0126] When performing the second attention enhancement, various machine learning or deep learning algorithms can be used, such as attention mechanism, convolutional neural network, etc. These algorithms can automatically learn and identify the key information in the multimodal entity semantic samples and enhance them so as to better utilize this information in the subsequent algorithm weight optimization.

[0127] Secondly, the second integration processing of the algorithm weights and the second multimodal attention enhancement semantics is another key step of the technical solution. In the initial radar data matching analysis algorithm, the algorithm weights determine the importance of different data features in the matching analysis. However, these algorithm weights are often set based on experience or experiments during algorithm design, and may not be fully adapted to actual application scenarios. Therefore, through the second integration processing, the key information in the second multimodal attention enhancement semantics can be fused with the algorithm weights, so that the algorithm weights can more accurately reflect the importance of data features in actual application scenarios.

[0128] When performing the second integration processing, various integration learning methods can be used, such as weighted average, voting mechanism, stacking generalization, etc. These methods can effectively combine the key information in the second multi-modal attention reinforcement semantics with the algorithm weights, generate new algorithm weights, and significantly improve the performance of the pre-debugging radar data matching analysis algorithm.

[0129] For example, the second integrated processing process may be to weightedly fuse each key information in the second multimodal attention enhancement semantics with the corresponding features in the algorithm weights. In the weighted fusion process, different weights may be given according to the importance of the key information to ensure that the more important information is more reflected in the algorithm weights. Through such processing, the pre-debugging radar data matching analysis algorithm can more accurately identify key data features and perform effective matching analysis when facing actual radar data.

[0130] Finally, through the second attention enhancement and second integration processing, the radar data analysis system can obtain the pre-debugging radar data matching analysis algorithm. While retaining the basic framework and principles of the initial radar data matching analysis algorithm, the algorithm weights are optimized by introducing key information in the multimodal entity semantic samples. Such optimization makes the pre-debugging radar data matching analysis algorithm have higher accuracy and efficiency in practical applications, and can better meet the needs of radar data matching analysis.

[0131] In this way, by making full use of the rich information in the multi-modal entity semantic samples, the algorithm weights in the initial radar data matching analysis algorithm are effectively optimized. The optimized pre-debugging radar data matching analysis algorithm has higher performance and stability in practical applications, and can provide more accurate and reliable results for radar data matching analysis. The proposal and implementation of this technical solution not only improves the technical level of radar data matching analysis, but also provides useful reference and reference for subsequent related research and applications.

[0132] In other independent embodiments, the pre-debugging radar data matching analysis algorithm is debugged based on the multimodal entity semantic sample to obtain the target radar data matching analysis algorithm, including: integrating attention enhancement on the multimodal linkage semantics and the multimodal entity semantic samples to obtain the matching training probability of the linkage radar scanning data for the target historical airborne radar three-dimensional scanning data; obtaining the matching prior probability between the linkage radar scanning data and the target historical airborne radar three-dimensional scanning data; determining the target data matching error of the pre-debugging radar data matching analysis algorithm based on the matching training probability and the matching prior probability; and debugging the pre-debugging radar data matching analysis algorithm based on the target data matching error to obtain the target radar data matching analysis algorithm.

[0133] In this embodiment, the core of the technical solution is to further debug the pre-debugged radar data matching analysis algorithm using multi-mode entity semantic samples to obtain the target radar data matching analysis algorithm. This process involves the in-depth integration of multi-mode linkage semantics and multi-mode entity semantic samples, as well as the comprehensive consideration of matching training probability and matching prior probability, aiming to ensure the accuracy and reliability of the radar data matching analysis algorithm.

[0134] First, it is an important step of this technical solution to integrate and strengthen the multi-mode linkage semantics and multi-mode entity semantic samples. Multi-mode linkage semantics contains rich content related to dynamic information such as time, events, and behaviors, while multi-mode entity semantic samples are the result of comprehensive processing of airborne radar scanning data, multi-mode linkage semantics, and regional element semantics. Through integrated attention strengthening, the key information of the two can be effectively fused to obtain the matching training probability of linkage radar scanning data for the target historical airborne radar 3D scanning data. The purpose of this step is to highlight the most critical information for radar data matching analysis, and provide strong support for the subsequent determination of target data matching errors.

[0135] When performing integrated attention enhancement, the attention mechanism in deep learning can be used to automatically learn and identify key information in multi-modal linkage semantics and multi-modal entity semantics samples through training models, and then fuse them. In this way, the radar data analysis system can obtain more accurate and comprehensive matching training probabilities, laying a solid foundation for subsequent steps.

[0136] Secondly, obtaining the matching prior probability between the linkage radar scan data and the target's historical airborne radar 3D scan data is another key step in the technical solution. The matching prior probability is obtained based on historical data and empirical knowledge, which reflects the matching possibility between the linkage radar scan data and the target's historical airborne radar 3D scan data without any additional information. The purpose of this step is to provide a benchmark for the subsequent debugging process so that the target data matching error can be determined more accurately.

[0137] When obtaining the prior probability of matching, the radar data analysis system can use the matching results and statistical data in the historical data to calculate the prior probability of matching through probability statistics. In this way, the radar data analysis system can obtain a relatively accurate and reliable benchmark value, providing strong support for subsequent steps.

[0138] Then, based on the matching training probability and matching prior probability, the radar data analysis system can determine the target data matching error of the pre-debugging radar data matching analysis algorithm. The target data matching error is an important indicator to measure the performance of the algorithm, which reflects the accuracy and reliability of the algorithm in practical applications. By comprehensively considering the matching training probability and matching prior probability, the radar data analysis system can obtain a more comprehensive and accurate target data matching error, providing a strong basis for subsequent steps.

[0139] When determining the target data matching error, the radar data analysis system can use various error calculation methods, such as mean square error, absolute error, etc. These methods can be selected according to the actual application scenarios and requirements to ensure that the obtained target data matching error has practical significance and application value.

[0140] Finally, based on the target data matching error, the radar data analysis system can further debug the pre-debugging radar data matching analysis algorithm to obtain the target radar data matching analysis algorithm. The purpose of this step is to further improve the performance of the algorithm in practical applications by adjusting the parameters and strategies in the algorithm. Through continuous debugging and optimization, the radar data analysis system can obtain a more accurate, reliable and efficient radar data matching analysis algorithm, providing strong support for practical applications.

[0141] During debugging, the radar data analysis system can use various optimization algorithms and machine learning techniques, such as gradient descent, random forest, etc. These methods can automatically learn and adjust according to the target data matching error to find the optimal algorithm parameters and strategies. Through such a debugging process, the radar data analysis system can obtain a target radar data matching analysis algorithm with better performance.

[0142] In this way, by making full use of the rich information in the multi-mode linkage semantics and multi-mode entity semantics samples, the pre-debugging radar data matching analysis algorithm was further debugged and optimized. By integrating attention enhancement, obtaining matching prior probability, determining target data matching error, and error-based debugging, the radar data analysis system has obtained a target radar data matching analysis algorithm with more excellent performance, accuracy and reliability. The proposal and implementation of this technical solution not only improves the technical level of radar data matching analysis, but also provides useful reference and reference for subsequent related research and applications.

[0143] In some other independent embodiments, the multimodal entity semantic samples include multimodal entity semantic samples of associated entity clusters in a verification sample batch; the pre-debugging radar data matching analysis algorithm is debugged based on the target data matching error to obtain a target radar data matching analysis algorithm, including: based on the target data matching error, the algorithm weights in the pre-debugging radar data matching analysis algorithm are corrected to obtain an intermediate radar data matching analysis algorithm; multimodal entity semantic samples in a test sample batch are obtained, and based on the multimodal entity semantic samples in the test sample batch, a first target data matching error of the intermediate radar data matching analysis algorithm is determined; if the first target data matching error meets the set error index, the intermediate radar data matching analysis algorithm is used as the target radar data matching analysis algorithm; if the first target data matching error does not meet the set error index, the algorithm module weights in the initial radar data matching analysis algorithm are corrected as a whole based on the first target data matching error, and the step of performing multimodal semantic mining on the knowledge vector entity and the associated vector entity respectively is jumped to.

[0144] It can be understood that the core of this embodiment is to use multi-mode entity semantic samples to deeply debug and optimize the pre-debugging radar data matching analysis algorithm to finally obtain the target radar data matching analysis algorithm. This process not only involves fine correction of the algorithm weights, but also includes full use of multi-mode entity semantic samples in the test sample batch, and overall adjustment of the algorithm module weights, aiming to ensure the excellent performance and high accuracy of the radar data matching analysis algorithm.

[0145] First, multimodal entity semantic samples play a crucial role in this technical solution. They not only contain rich information about associated entity clusters in the verification sample batch, but also contain deep connections between multimodal linkage semantics, regional element semantics, and airborne radar scanning data. These samples provide valuable reference and basis for debugging and optimizing the algorithm.

[0146] Based on the target data matching error, the radar data analysis system corrects the algorithm weights in the pre-debugging radar data matching analysis algorithm, which is the first step of the technical solution. Algorithm weights play a pivotal role in the radar data matching analysis algorithm, which determines the importance and influence of different data features in the matching process. By correcting the algorithm weights, the radar data analysis system can make the algorithm pay more attention to those data features that have a significant impact on the matching results, thereby improving the accuracy and robustness of the algorithm. The completion of this step indicates that the radar data analysis system has obtained the intermediate radar data matching analysis algorithm, which is an important step towards the target algorithm.

[0147] However, simply correcting the algorithm weights is not enough to ensure that the target algorithm obtained by the radar data analysis system has the best performance. Therefore, the radar data analysis system needs to further obtain multi-modal entity semantic samples in the test sample batch and determine the first target data matching error of the intermediate radar data matching analysis algorithm based on these samples. The purpose of this step is to verify the performance of the intermediate algorithm through actual test data and evaluate whether it has reached the error index set by the radar data analysis system.

[0148] If the first target data matching error meets the set error index, then the radar data analysis system can be confident that the intermediate radar data matching analysis algorithm has excellent performance and accuracy. At this point, the radar data analysis system can use the intermediate algorithm as the target radar data matching analysis algorithm and apply it to the actual radar data matching task. This is undoubtedly the best result expected by the radar data analysis system, because it means that the technical solution of the radar data analysis system has been successful and a target algorithm with excellent performance has been obtained.

[0149] However, if the first target data matching error does not meet the set error index, the radar data analysis system needs to further adjust and optimize the technical solution. Specifically, the radar data analysis system will make an overall correction to the algorithm module weights in the initial radar data matching analysis algorithm based on the first target data matching error. The purpose of this step is to improve the overall performance of the algorithm by adjusting the relative importance between algorithm modules. By making an overall correction to the algorithm module weights, the radar data analysis system can make the algorithm more flexibly adapt to different matching tasks and data features, thereby improving its accuracy and robustness.

[0150] After completing the overall correction of the algorithm module weights, the radar data analysis system needs to jump to the step of multi-modal semantic mining of knowledge vector entities and association vector entities. This step is an important part of the technical solution. It extracts more useful information and features through in-depth analysis and mining of knowledge vector entities and association vector entities. These information and features can provide strong support and guidance for further debugging and optimization of the algorithm. Through multi-modal semantic mining of knowledge vector entities and association vector entities, the radar data analysis system can discover more potential data associations and matching rules, thereby providing new ideas and directions for algorithm improvement.

[0151] With this design, the pre-debugging radar data matching analysis algorithm is deeply debugged and optimized by making full use of the data in the multi-modal entity semantic samples and test sample batches. Through the correction of algorithm weights, the overall adjustment of algorithm module weights, and the multi-modal semantic mining of knowledge vector entities and associated vector entities, the radar data analysis system obtains a target radar data matching analysis algorithm with excellent performance and high accuracy.

[0152] Exemplarily, the following are exemplary formulas for the above key steps.

[0153] (1) Algorithm weight correction formula:

[0154] W_new = Optimize(W_old, E)

[0155] Among them, W_old represents the algorithm weight before correction, W_new represents the algorithm weight after correction, E represents the target data matching error, and Optimize is a generalized optimization function used to adjust the weight according to the error.

[0156] (2) Formula for determining the first target data matching error:

[0157] E_first=Evaluate(A_intermediate,S_test)

[0158] Among them, A_intermediate represents the intermediate radar data matching analysis algorithm, S_test represents the multimodal entity semantic samples in the test sample batch, Evaluate is an evaluation function used to calculate the error generated by the algorithm when processing the test sample, and E_first represents the first target data matching error obtained.

[0159] (3) Overall correction formula of algorithm module weight:

[0160] M_new=AdjustModules(M_old,E_first)

[0161] Among them, M_old represents the algorithm module weight before correction, M_new represents the algorithm module weight after correction, E_first represents the first target data matching error, and AdjustModules is a module weight adjustment function used to adjust the weights between algorithm modules according to the error.

[0162] (4) Multimodal semantic mining steps:

[0163] PerformSemanticMining(KnowledgeVectors, AssociationVectors)

[0164] Among them, PerformSemanticMining is a function that performs multimodal semantic mining, KnowledgeVectors represents knowledge vector entities, and AssociationVectors represents association vector entities. This step aims to discover new information and features by deeply analyzing the semantic relationships between entities.

[0165] Further, Figure 2 Schematic diagram of the structure of a radar data analysis system 200 provided in an embodiment of the present application. Figure 2 The radar data analysis system 200 shown includes a processor 210, which can call and run a computer program from a memory to implement the method in the embodiment of the present application.

[0166] Alternatively, if Figure 2 As shown, the radar data analysis system 200 may further include a memory 230. The processor 210 may call and run a computer program from the memory 230 to implement the method in the embodiment of the present application.

[0167] The memory 230 may be a separate device independent of the processor 210 , or may be integrated into the processor 210 .

[0168] Alternatively, if Figure 2As shown, the radar data analysis system 200 may further include a transceiver 220, and the processor 210 may control the transceiver 220 to interact with other devices, specifically, may send information or data to other devices, or receive information or data sent by other devices.

[0169] Optionally, the radar data analysis system 200 may implement the corresponding processes corresponding to the storage engine or components in the storage engine (such as a processing module) or a device deployed with a storage engine in each method of the embodiments of the present application, which will not be described in detail here for the sake of brevity.

[0170] It should be understood that the processor of the embodiment of the present application may be an integrated circuit chip with signal processing capabilities.

[0171] It is understood that the memory in the embodiments of the present application may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memory. It should be noted that the memory of the systems and methods described herein is intended to include but is not limited to suitable types of memory.

[0172] Based on the above, a readable storage medium is provided, on which a program or instruction is stored, and when the program or instruction is executed by a processor, the steps of the above method are implemented.

[0173] It should be noted that, in this article, the terms "comprise", "include" or any other variant thereof are intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise one..." do not exclude the presence of other identical elements in the process, method, article or device including the element. In addition, it should be pointed out that the scope of the methods and devices in the embodiments of the present application is not limited to performing functions in the order shown or discussed, and may also include performing functions in a substantially simultaneous manner or in reverse order according to the functions involved, for example, the described method may be performed in an order different from that described, and various steps may also be added, omitted, or combined. In addition, the features described with reference to certain examples may be combined in other examples.

[0174] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a computer software product, which is stored in a storage medium (such as ROM / RAM, a magnetic disk, or an optical disk), and includes a number of instructions for a terminal (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods described in each embodiment of the present application.

[0175] The embodiments of the present application are described above in conjunction with the accompanying drawings, but the present application is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present application, ordinary technicians in this field can also make many forms without departing from the purpose of the present application and the scope of protection of the present application, all of which are within the protection of the present application.

Claims

1. A method for analyzing radar data onboard a drone, characterized in that: The method is applied to a radar data analysis system, and the method comprises: Acquire a target knowledge vector entity of the airborne radar three-dimensional scanning data to be processed in the target spatiotemporal knowledge relationship space and a target association vector entity corresponding to the target knowledge vector entity, wherein the target association vector entity includes a target linkage entity corresponding to the target linkage radar scanning data involved in the airborne radar three-dimensional scanning data to be processed and a target area element entity corresponding to the surface area element of the airborne radar three-dimensional scanning data to be processed; Performing multimodal semantic mining on the target knowledge vector entity and the target association vector entity to obtain original multimodal entity semantics corresponding to the target knowledge vector entity, target multimodal linkage semantics corresponding to the target linkage entity, and target regional element semantics corresponding to the target regional element entity; Integrate the original multi-mode entity semantics, the target multi-mode linkage semantics and the target area element semantics to obtain the target multi-mode entity semantics of the airborne radar three-dimensional scanning data to be processed; Obtaining the multimodal entity semantics corresponding to the environment monitoring platform server in the target spatiotemporal knowledge relationship space; Determining a target matching probability of the environmental monitoring platform server for the airborne radar three-dimensional scanning data to be processed based on the target multi-mode entity semantics and the multi-mode entity semantics corresponding to the environmental monitoring platform server; The airborne radar three-dimensional scanning data to be processed with a target matching probability greater than a set probability is sent to the environment monitoring platform server.

2. The method according to claim 1, characterized in that The method is implemented by a target radar data matching analysis algorithm, and the debugging steps of the target radar data matching analysis algorithm include: Acquire a spatiotemporal knowledge relationship space generated according to an environmental monitoring log report in an environmental monitoring database, wherein the spatiotemporal knowledge relationship space includes a plurality of mapping entities, each of the mapping entities corresponds to the environmental monitoring log report, and the environmental monitoring log report includes historical airborne radar three-dimensional scanning data and surface area elements of the historical airborne radar three-dimensional scanning data; Determine, from the spatiotemporal knowledge relationship space, a knowledge vector entity corresponding to the target historical airborne radar three-dimensional scanning data in the historical airborne radar three-dimensional scanning data and an associated vector entity of the knowledge vector entity, wherein the associated vector entity includes a linkage entity corresponding to the linkage radar scanning data involved in the target historical airborne radar three-dimensional scanning data and an area element entity corresponding to the surface area element of the target historical airborne radar three-dimensional scanning data; Performing multi-mode semantic mining on the knowledge vector entity and the association vector entity respectively to obtain the airborne radar scanning data semantics corresponding to the knowledge vector entity, the multi-mode linkage semantics corresponding to the linkage entity, and the surface regional element semantics corresponding to the regional element entity; Integrate the semantics of the airborne radar scanning data, the multi-mode linkage semantics and the surface area element semantics to obtain a multi-mode entity semantic sample of the target historical airborne radar three-dimensional scanning data; Based on the multi-mode entity semantic sample, the algorithm weights in the initial radar data matching analysis algorithm are optimized to obtain a pre-debugging radar data matching analysis algorithm; Based on the multi-mode entity semantic sample, the pre-debugging radar data matching analysis algorithm is debugged to obtain a target radar data matching analysis algorithm.

3. The method according to claim 2, characterized in that The step of determining, from the spatiotemporal knowledge relationship space, a knowledge vector entity corresponding to the target historical airborne radar three-dimensional scanning data in the historical airborne radar three-dimensional scanning data and an associated vector entity of the knowledge vector entity comprises: Acquire a plurality of knowledge transfer strategies, wherein the knowledge transfer strategies include different entity index tags; Based on the different entity index tags, the knowledge vector entity corresponding to the target historical airborne radar three-dimensional scanning data in the historical airborne radar three-dimensional scanning data and the associated vector entity of the knowledge vector entity are determined from the spatiotemporal knowledge relationship space.

4. The method according to claim 3, characterized in that The different entity index tags include a head entity index tag and a preceding and following entity index tags; the determining, based on the different entity index tags, from the spatiotemporal knowledge relationship space, a knowledge vector entity corresponding to the target historical airborne radar three-dimensional scanning data in the historical airborne radar three-dimensional scanning data and an associated vector entity of the knowledge vector entity, including: Determine a mapping entity corresponding to the head entity index tag from the spatiotemporal knowledge relationship space, and obtain a knowledge vector entity corresponding to the historical airborne radar three-dimensional scanning data of the target; Obtaining an original association vector entity corresponding to the knowledge vector entity; The original relevance vector entity corresponding to the preceding and following entity index labels is determined from the original relevance vector entity to obtain the relevance vector entity of the knowledge vector entity.

5. The method according to claim 2, characterized in that: The association vector entity includes multiple layers of association vector entities, and the linkage entity and the regional element entity are first-layer association vector entities of the knowledge vector entity; The integrating processing of the semantics of the airborne radar scanning data, the multi-mode linkage semantics and the semantics of the surface area elements to obtain a multi-mode entity semantic sample of the target historical airborne radar three-dimensional scanning data includes: Determine a bottom layer relevance vector entity of the knowledge vector entity from the multi-layer relevance vector entity, and determine a previous layer relevance vector entity of the bottom layer relevance vector entity from the multi-layer relevance vector entity; Integrate the multimodal entity semantics of the bottom layer association vector entity and the multimodal entity semantics of the upper layer association vector entity to obtain a multimodal entity semantic sample of the upper layer association vector entity; If the upper layer association vector entity is not the first layer association vector entity, the multimodal entity semantic sample of the upper layer association vector entity is used as the multimodal entity semantics of the upper layer association vector entity, the upper layer association vector entity is used as the bottom layer association vector entity, and the process jumps to the step of determining the upper layer association vector entity of the bottom layer association vector entity from the multiple layers of association vector entities; If the upper layer association vector entity is the first layer association vector entity, the multimodal entity semantic sample of the upper layer association vector entity is used as the multimodal linkage semantics or the surface area element semantics; The semantics of the airborne radar scanning data, the multi-mode linkage semantics and the surface area element semantics are integrated and processed to obtain a multi-mode entity semantic sample of the target historical airborne radar three-dimensional scanning data.

6. The method according to claim 5, characterized in that The step of determining the bottom layer association vector entity of the knowledge vector entity from the multi-layer association vector entity, and determining the upper layer association vector entity of the bottom layer association vector entity from the multi-layer association vector entity, comprises: Acquire a knowledge transfer strategy, wherein the knowledge transfer strategy includes a preceding and following entity index tag; Based on the preceding and following entity index tags, a bottom-level association vector entity of the knowledge vector entity is determined from the multi-level association vector entities, and an upper-level association vector entity of the bottom-level association vector entity is determined from the multi-level association vector entities.

7. The method according to claim 5, characterized in that The integrating processing of the multimodal entity semantics of the bottom layer association vector entity and the multimodal entity semantics of the upper layer association vector entity to obtain a multimodal entity semantic sample of the upper layer association vector entity includes: Obtaining a correlation between the bottom layer association vector entity and the upper layer association vector entity, wherein the correlation represents an influence coefficient of the bottom layer association vector entity on the upper layer association vector entity; Based on the correlation, the multimodal entity semantics of the underlying association vector entity is updated to obtain updated multimodal entity semantics of the underlying association vector entity; The updated multimodal entity semantics and the multimodal entity semantics of the upper-layer association vector entity are integrated to obtain a multimodal entity semantic sample of the upper-layer association vector entity.

8. The method according to claim 7, characterized in that The obtaining of the correlation between the bottom layer association vector entity and the upper layer association vector entity includes: Determining a feature commonality value between the upper layer association vector entity and the lower layer association vector entity based on the multimodal entity semantics of the lower layer association vector entity and the multimodal entity semantics of the upper layer association vector entity; Based on the feature commonality value, the correlation between the bottom layer relevance vector entity and the upper layer relevance vector entity is determined.

9. The method according to claim 2, characterized in that: The integrating processing of the semantics of the airborne radar scanning data, the multi-mode linkage semantics and the semantics of the surface area elements to obtain a multi-mode entity semantic sample of the target historical airborne radar three-dimensional scanning data includes: Integrate the semantics of the airborne radar scanning data and the multi-mode linkage semantics to obtain a first multi-mode entity semantic sample of the historical airborne radar three-dimensional scanning data of the target; Integrate the semantics of the airborne radar scanning data and the semantics of the surface area elements to obtain a second multi-mode entity semantic sample of the historical airborne radar three-dimensional scanning data of the target; Determining a multimodal entity semantic sample of the target historical airborne radar three-dimensional scanning data based on the first multimodal entity semantic sample and the second multimodal entity semantic sample; The determining, based on the first multimodal entity semantic sample and the second multimodal entity semantic sample, the multimodal entity semantic sample of the target historical airborne radar three-dimensional scanning data includes: Obtaining a first correlation corresponding to the first multimodal entity semantic example and a second correlation corresponding to the second multimodal entity semantic example; Based on the first correlation, updating the first multimodal entity semantic sample to obtain an updated first multimodal entity semantic sample; Based on the second correlation, updating the second multimodal entity semantic sample to obtain an updated second multimodal entity semantic sample; Determining a multimodal entity semantic sample of the target historical airborne radar three-dimensional scanning data based on the updated first multimodal entity semantic sample and the updated second multimodal entity semantic sample; The obtaining of a first correlation corresponding to the first multimodal entity semantic sample and a second correlation corresponding to the second multimodal entity semantic sample includes: Performing a first integration process on the first multimodal entity semantic sample and the second multimodal entity semantic sample to obtain integrated multimodal entity semantics; Performing first attention enhancement on the integrated multimodal entity semantics to obtain first multimodal attention enhanced semantics; A first correlation corresponding to the first multimodal entity semantic sample and a second correlation corresponding to the second multimodal entity semantic sample are determined from the first multimodal attention-enhanced semantics.

10. A radar data analysis system, characterized in that: The method comprises at least one processor and a memory; the memory stores computer-executable instructions; the at least one processor executes the computer-executable instructions stored in the memory, so that the at least one processor executes the method according to any one of claims 1 to 9.

Images