Low-altitude defense scheme adjustment method and device based on multi-modal feature fusion

Abstract

Embodiments of the present disclosure disclose a low-altitude defense scheme adjustment method and device based on multi-modal feature fusion. A specific implementation of the method includes: converting a collected low-altitude related data set for a low-altitude target detection scene into a multi-modal feature data set; correlatively encoding the multi-modal feature data set to generate a flight object perception feature map; generating a threat level sequence according to a flight object time sequence feature sequence; scene parameterizing the threat level sequence to generate a parameterized flight object scene model; determining a target low-altitude defense execution scheme based on the parameterized flight object scene model; and in response to determining that an execution deviation value corresponding to the target low-altitude defense execution scheme exceeds a preset deviation threshold, dynamically adjusting the target low-altitude defense execution scheme. This implementation reduces the false positive rate of low-altitude flight object data, can adapt to dynamically changing threat behaviors and environmental conditions, and timely adjusts the low-altitude defense scheme.

Description

Technical Field

[0001] The embodiments disclosed herein relate to the field of computer technology, and specifically to a method and apparatus for adjusting low-altitude defense schemes based on multimodal feature fusion. Background Technology

[0002] Low-altitude defense scheme adjustment based on multimodal feature fusion is a technique for adjusting low-altitude defense schemes. Currently, the common methods for adjusting low-altitude defense schemes are: using a single sensor to collect data on low-altitude flying objects to adjust the resulting low-altitude defense scheme, or relying on manual judgment to adjust the low-altitude defense scheme. For example, using radar alone to detect low-altitude flying objects in multiple directions, speeds, and altitudes.

[0003] However, when adjusting low-altitude defense schemes using the above methods, the following technical problems often arise:

[0004] When a single sensor collects data on low-altitude flying objects in three-dimensional space, the detection blind zone is large, which can easily lead to missed targets or interrupted trajectories, resulting in a high false alarm rate. Relying on human experience often results in slow response speed, difficulty in adapting to dynamically changing threat behaviors and environmental conditions, and difficulty in timely adjusting low-altitude defense strategies in scenarios with multiple targets operating concurrently. Summary of the Invention

[0005] The summary portion of this disclosure is intended to provide a brief overview of the concepts, which will be described in detail in the detailed description portion. This summary portion is not intended to identify key or essential features of the claimed technical solutions, nor is it intended to limit the scope of the claimed technical solutions.

[0006] Some embodiments of this disclosure propose a method and apparatus for adjusting low-altitude defense schemes based on multimodal feature fusion to solve the technical problems mentioned in the background section above.

[0007] Firstly, some embodiments of this disclosure provide a method for adjusting a low-altitude defense scheme based on multimodal feature fusion. The method includes: converting a collected low-altitude related dataset for a low-altitude target detection scenario into a multimodal feature dataset, wherein the low-altitude target detection scenario represents a scenario where an undetected low-altitude flying object exists in a low-altitude area within a preset range, and the low-altitude related dataset is a dataset related to low-altitude flying objects collected in real time by multiple pre-deployed sensors; performing association encoding on the multimodal feature dataset to generate a flying object perception feature map, wherein each node of the flying object perception feature map represents flying object feature data at different times, and each edge of the perception feature map represents the flight trajectory of the flying object; and processing the flying object perception feature map... The system extracts temporal node sequences from the spectrum to generate a temporal feature sequence of the flying object, where the temporal feature sequence represents the flight state changes of the same flying object at different times. Based on the temporal feature sequence, a threat level sequence is generated, where the threat level sequence represents the threat level of the flying object at different trajectories over time. The threat level sequence is then parameterized to generate a parameterized flying object scenario model. Based on the parameterized flying object scenario model, a target low-altitude defense execution plan is determined, where the target low-altitude defense execution plan is the execution plan for defending against the low-altitude flying object. In response to determining that the execution deviation value corresponding to the target low-altitude defense execution plan exceeds a preset deviation threshold, the target low-altitude defense execution plan is dynamically adjusted.

[0008] Secondly, some embodiments of this disclosure provide a low-altitude defense scheme adjustment device based on multimodal feature fusion. The device includes: a conversion unit configured to convert a collected low-altitude related dataset for a low-altitude target detection scenario into a multimodal feature dataset, wherein the low-altitude target detection scenario represents a scenario where an undetected low-altitude flying object exists in a low-altitude area within a preset range, and the low-altitude related dataset is a dataset related to low-altitude flying objects collected in real time by multiple pre-deployed sensors; an encoding unit configured to perform associative encoding on the multimodal feature dataset to generate a flying object perception feature map, wherein each node of the flying object perception feature map represents flying object feature data at different times, and each edge of the perception feature map represents the flight trajectory of the flying object; and an extraction unit configured to perform temporal node extraction on the flying object perception feature map. Sequence extraction is performed to generate a temporal feature sequence of an object, wherein the temporal feature sequence represents the flight state changes of the same object at different times; a generation unit is configured to generate a threat level sequence based on the temporal feature sequence of the object, wherein the threat level sequence represents the threat level of the object at different trajectories over time; a parameterization unit is configured to parameterize the threat level sequence to generate a parameterized object scenario model; a determination unit is configured to determine a target low-altitude defense execution plan based on the parameterized object scenario model, wherein the target low-altitude defense execution plan is an execution plan for defending against the low-altitude object; and a dynamic adjustment unit is configured to dynamically adjust the target low-altitude defense execution plan in response to the determination that the execution deviation value corresponding to the target low-altitude defense execution plan exceeds a preset deviation threshold.

[0009] Thirdly, some embodiments of this disclosure provide an electronic device, including: one or more processors; and a storage device having one or more programs stored thereon, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the method described in any implementation of the first aspect above.

[0010] Fourthly, some embodiments of this disclosure provide a computer-readable medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the method described in any of the implementations of the first aspect above.

[0011] The various embodiments disclosed herein have the following beneficial effects: The low-altitude defense scheme adjustment method based on multimodal feature fusion, as described in some embodiments of this disclosure, reduces the false alarm rate of low-altitude flying object data, adapts to dynamically changing threat behaviors and environmental conditions, and allows for timely adjustment of the low-altitude defense scheme. Specifically, the high false alarm rate and difficulty in adapting to dynamically changing threat behaviors and environmental conditions are due to the following: When a single sensor collects data on low-altitude flying objects in three-dimensional space, the detection blind zone is large, easily leading to missed target detection or trajectory interruption, resulting in a high false alarm rate. Relying on human experience often results in a slow response speed, making it difficult to adapt to dynamically changing threat behaviors and environmental conditions, and making it difficult to adjust the low-altitude defense scheme in a timely manner in scenarios with multiple concurrent targets. Based on this, the low-altitude defense scheme adjustment method based on multimodal feature fusion in some embodiments of this disclosure firstly transforms the collected low-altitude related dataset for low-altitude target detection scenarios into a multimodal feature dataset. The low-altitude target detection scenario represents a scenario where undetected low-altitude flying objects exist in a preset range of low-altitude areas. The low-altitude related dataset is a dataset related to low-altitude flying objects collected in real time by multiple pre-deployed sensors. This enables complementary fusion and feature enhancement of multi-source heterogeneous data, effectively overcoming the detection blind spots and false alarms of single sensors in three-dimensional space. By fusing data collected from multiple sensors such as radar, photoelectric, infrared, and radio sensors, a more comprehensive and accurate perception capability of low-altitude flying objects is formed, reducing the false alarm rate of low-altitude flying object data. Then, the multimodal feature dataset is correlated and encoded to generate a flying object perception feature map. Each node of the flying object perception feature map represents flying object feature data at different times, and each edge of the perception feature map represents the flying object's flight trajectory. This enables the construction of a visual map of flying objects, effectively solving the problems of target miss detection or trajectory interruption. Next, the temporal node sequence of the aforementioned object perception feature map is extracted to generate an object temporal feature sequence, which represents the flight state changes of the same object at different times. This allows for the extraction and characterization of the continuous behavioral evolution of the same object, forming its complete flight state time series. Secondly, based on the aforementioned object temporal feature sequence, a threat level sequence is generated, representing the threat level of the object at different trajectories over time. This can replace subjective judgments relying on human experience, providing objective data. Then, the threat level sequence is parameterized to generate a parameterized object scenario model. This allows the dynamic threat situation to be abstracted into a simulable digital scenario, providing a foundation for subsequent processing. Finally, based on the parameterized object scenario model, a target low-altitude defense execution plan is determined, which is the execution plan for defending against the aforementioned low-altitude objects.Therefore, target low-altitude defense execution plans can be automatically generated based on digital scenarios, overcoming the limitations of manually formulated plans, such as slow response, poor adaptability, and difficulty in handling multiple concurrent targets. Finally, in response to the determination that the execution deviation value corresponding to the above target low-altitude defense execution plan exceeds a preset deviation threshold, the above target low-altitude defense execution plan is dynamically adjusted. This allows for adaptation to dynamically changing threat behaviors and environmental conditions, enabling timely adjustments to the low-altitude defense plan even in scenarios with multiple concurrent targets. Thus, it reduces the false alarm rate of low-altitude flying object data, adapts to dynamically changing threat behaviors and environmental conditions, and allows for timely adjustments to the low-altitude defense plan. Attached Figure Description

[0012] The above and other features, advantages, and aspects of the embodiments of this disclosure will become more apparent from the accompanying drawings and the following detailed description. Throughout the drawings, the same or similar reference numerals denote the same or similar elements. It should be understood that the drawings are schematic, and elements are not necessarily drawn to scale.

[0013] Figure 1 This is a schematic diagram of an application scenario of the low-altitude defense scheme adjustment method based on multimodal feature fusion, which is one of the embodiments of this disclosure;

[0014] Figure 2 This is a flowchart of some embodiments of the low-altitude defense scheme adjustment method based on multimodal feature fusion according to the present disclosure;

[0015] Figure 3 This is a schematic diagram of the structure of some embodiments of the low-altitude defense scheme adjustment device based on multimodal feature fusion according to the present disclosure;

[0016] Figure 4 This is a schematic diagram of the structure of an electronic device suitable for implementing some embodiments of the present disclosure;

[0017] Figure 5 These are flight object perception feature maps based on some embodiments of the low-altitude defense scheme adjustment method based on multimodal feature fusion disclosed herein;

[0018] Figure 6 This is a threat level sequence diagram based on some embodiments of the low-altitude defense scheme adjustment method based on multimodal feature fusion disclosed herein. Detailed Implementation

[0019] Embodiments of this disclosure will now be described in more detail with reference to the accompanying drawings. While some embodiments of this disclosure are shown in the drawings, it should be understood that this disclosure can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of this disclosure. It should be understood that the accompanying drawings and embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of protection of this disclosure.

[0020] It should also be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings. Unless otherwise specified, the embodiments and features described in this disclosure can be combined with each other.

[0021] It should be noted that the concepts of "first" and "second" mentioned in this disclosure are used only to distinguish different devices, modules or units, and are not used to limit the order of functions performed by these devices, modules or units or their interdependencies.

[0022] It should be noted that the terms "a" and "a plurality of" used in this disclosure are illustrative rather than restrictive, and those skilled in the art should understand that, unless otherwise expressly indicated in the context, they should be understood as "one or more".

[0023] The names of messages or information exchanged between multiple devices in the embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of such messages or information.

[0024] This disclosure will now be described in detail with reference to the accompanying drawings and embodiments.

[0025] Figure 1 This is a schematic diagram of an application scenario of the low-altitude defense scheme adjustment method based on multimodal feature fusion, which is one of the embodiments of this disclosure.

[0026] exist Figure 1 In the application scenario, firstly, the collected low-altitude related dataset for low-altitude target detection scenario 100 is transformed into a multimodal feature dataset. The aforementioned low-altitude target detection scenario represents a low-altitude region within a preset range (e.g., low-altitude target detection scenario 100). Figure 1 The scenario where an undetected low-altitude flying object 101 exists within the square frame (as defined in the image), the aforementioned low-altitude related dataset is generated by pre-deployed multiple sensors (such as...). Figure 1 The dataset related to low-altitude flying objects is collected in real time by an infrared thermal imager 102, a radar 103, an electro-optical camera 104, and a radio detection device 105. The position of the low-altitude flying object can be the coordinate position with the ground center O as the origin. Where x is the horizontal axis, y is the vertical axis, and z is the vertical axis. Figure 1 The rhombuses in the diagram represent the ground covered by the low-altitude region. It should be understood that... Figure 1 The number of sensors and low-altitude flying objects in the system can be arbitrary, depending on the implementation requirements.

[0027] Continue to refer to Figure 2 The diagram illustrates a flow 200 of some embodiments of a low-altitude defense scheme adjustment method based on multimodal feature fusion according to the present disclosure. This low-altitude defense scheme adjustment method based on multimodal feature fusion includes the following steps:

[0028] Step 201: Convert the collected low-altitude related dataset for low-altitude target detection scenarios into a multimodal feature dataset.

[0029] In some embodiments, the execution entity (e.g., a computing device) of the low-altitude defense scheme adjustment method based on multimodal feature fusion can transform the collected low-altitude related dataset for low-altitude target detection scenarios into a multimodal feature dataset. The aforementioned low-altitude target detection scenario represents a scenario where undetected low-altitude flying objects exist in a predetermined low-altitude region, and the aforementioned low-altitude related dataset is a dataset related to low-altitude flying objects collected in real time by multiple pre-deployed sensors.

[0030] The low-altitude related data in the aforementioned low-altitude related dataset may include, but is not limited to, at least one of the following: the position, speed, and heading of low-altitude flying objects. The aforementioned low-altitude flying objects may be undetected or unauthorized flying objects within a preset altitude and range in a low-altitude region. For example, the aforementioned low-altitude flying objects may be birds or drones. The position of the aforementioned low-altitude flying objects may be a position with coordinates (100, 20, 30) with the center of the ground as the origin. The aforementioned speed may be 300 km / h. The heading may be due east. The aforementioned preset range may be a pre-defined area of ​​1 hectare. The aforementioned preset altitude may be a pre-defined altitude of 100 meters above the ground. The multimodal feature data in the aforementioned multimodal feature dataset may be feature data that integrates features from multiple sensors. For example, multiple sensors may include, but are not limited to: radar, photoelectric cameras, radio detection equipment, and infrared thermal imagers. The aforementioned multiple sensor data features may include: spatial dimension data features and physical dimension data features. For example, the aforementioned spatial dimension data features may be represented as the three-dimensional position coordinates of the flying objects detected by the sensors. The aforementioned physical dimension data characteristics can be represented as the thermal radiation characteristics of flying objects detected by sensors.

[0031] As an example, the aforementioned execution entity can preprocess the collected low-altitude related datasets for low-altitude target detection scenarios to generate processed low-altitude related datasets. Then, data fusion is performed on each processed low-altitude related data point in the aforementioned processed low-altitude related dataset to generate fused low-altitude data, resulting in a fused low-altitude dataset. This data fusion can be a combination; for example, combining the position, velocity, and heading of low-altitude flying objects included in the processed low-altitude related data into fused low-altitude data. Multimodal feature extraction is then performed on the aforementioned fused low-altitude dataset to generate a multimodal feature dataset. These multimodal features may include, but are not limited to: the three-dimensional position coordinates of flying objects, the shape features of flying objects, and the flight speed of flying objects. The aforementioned data preprocessing can be data cleaning, used to remove duplicate data from the low-altitude related datasets.

[0032] Optionally, the aforementioned implementing entity can convert the collected low-altitude related dataset for low-altitude target detection scenarios into a multimodal feature dataset through the following steps:

[0033] The first step is to align the timestamps of the aforementioned low-altitude related datasets to generate aligned low-altitude related datasets, wherein each aligned low-altitude related data in the aforementioned aligned low-altitude related datasets has a unified timestamp sequence.

[0034] As an example, the aforementioned execution entity can use PTP (Precise Time Protocol) to timestamp-align the aforementioned low-altitude related dataset to generate an aligned low-altitude related dataset. For example, an optical camera might capture 30 frames per second, a vibration sensor 100 times per second, and radar 10 times per second. Using the vibration sensor as a reference, the optical and radar data can be interpolated and aligned to 100 time points per second.

[0035] The second step involves performing spatial coordinate transformation on the aforementioned aligned low-altitude correlation dataset to generate a transformed low-altitude correlation dataset. In this transformed low-altitude correlation dataset, each transformed low-altitude correlation data point is mapped to the same spatial coordinate system.

[0036] As an example, the aforementioned implementing entity can transform the aligned low-altitude correlation dataset to a unified GIS coordinate system to generate a transformed low-altitude correlation dataset. For instance, images captured by a camera mounted on a 10-meter-high tower can be transformed using perspective to determine the ground position of each pixel in the image within a GIS coordinate system. This GIS coordinate system can be the WGS-84 coordinate system.

[0037] The third step is to perform spatiotemporal registration and fusion processing on the above-mentioned aligned low-altitude related dataset and the above-mentioned transformed low-altitude related dataset to generate a spatiotemporal aligned dataset. Each spatiotemporal aligned data in the above-mentioned spatiotemporal aligned dataset includes: the position data, velocity data and heading data of the flying object under the same spatiotemporal information.

[0038] As an example, the aforementioned execution entity can use a data association algorithm to match the aligned low-altitude related dataset with the transformed low-altitude related dataset, and then perform a weighted fusion of the successfully matched related datasets. Finally, the fused related dataset is determined as the spatiotemporally aligned dataset. The aforementioned data association algorithm can be the Nearest Neighbor Data Association (NNDA) algorithm. For example, for each aligned low-altitude related dataset, the nearest transformed low-altitude related dataset is selected as a matching candidate, and the related datasets corresponding to the matching candidate are fused to obtain the spatiotemporally aligned dataset. The fourth step involves performing multimodal feature extraction processing on the aforementioned spatiotemporally aligned dataset to generate a multimodal feature dataset.

[0039] As an example, the aforementioned execution entity can utilize spatial transformation-based methods in feature extraction to extract multimodal features from the aforementioned spatiotemporally aligned dataset, thereby generating a multimodal feature dataset. Here, spatial transformation-based methods in feature extraction refer to mathematical coordinate transformations, projections, mappings, etc. Multimodal features can include, but are not limited to, spatiotemporal features and motion features. For example, a multimodal feature dataset could be "Spatiotemporal features: [position X=1297.5, position Y=749.8, position Z=257.5, velocity=80.2], motion features: [acceleration=3.0, maneuverability index=0.8, trajectory curvature=0.1]".

[0040] Step 202: Associatively encode the above multimodal feature dataset to generate an object perception feature map.

[0041] In some embodiments, the execution entity may perform associative encoding on the multimodal feature dataset to generate an object perception feature map. Each node in the object perception feature map represents object feature data at different times, and each edge of the perception feature map represents the object's flight trajectory.

[0042] Here, the aforementioned flying object characteristic data may include, but is not limited to, the flying object's three-dimensional coordinate position and flight speed. The aforementioned flight trajectory may be a straight line trajectory or a curved trajectory.

[0043] As an example, the aforementioned execution entity can use a feature-level fusion algorithm to correlate and encode the multimodal feature dataset to generate an object perception feature map. This feature-level fusion algorithm can be used to integrate various multimodal feature data from the multimodal feature dataset to form a joint feature representation. Specifically, the feature-level fusion algorithm integrates, correlates, and encodes multimodal feature data from different sensors to form a more informative joint feature representation, i.e., an object perception feature map. Integration, correlation, and encoding are mathematical methods. For example, multimodal feature data from different sensors can be concatenated.

[0044] In the process of adopting technical solutions to address the problems mentioned in the background section, the following issues often arise:

[0045] Because relying on human experience makes it difficult to quantify the correlation strength of flying objects in time, space and feature dimensions, data points become isolated and cannot form a coherent flight trajectory. This may lead to interruptions in the flight trajectory, which could cause the same flying object to be identified as two independent targets by the system before and after the interruption, thus making the low-altitude defense scheme less reliable.

[0046] In response to the aforementioned technical problems, the following solution was adopted:

[0047] Optionally, the aforementioned execution entity can perform associative encoding on the multimodal feature dataset using the following steps to generate an object perception feature map:

[0048] The first step is to generate node identifiers for each multimodal feature data in the aforementioned multimodal feature dataset, thereby generating multimodal feature nodes with unique identifiers and obtaining a multimodal feature node set. These identifiers are generated based on a combination of temporal and spatial information.

[0049] Here, the time information mentioned above can be a timestamp, and the spatial information can be three-dimensional location coordinates. For example, the time information could be 10:30:20:15 AM on December 29, 2025. The spatial information could be (110, 25, 50).

[0050] As an example, the aforementioned execution entity can determine a node identifier for each multimodal feature data in the aforementioned multimodal feature dataset. For instance, the temporal information and spatial coordinates of the multimodal feature data can be combined as the node identifier. Node identifier = "timestamp" + "spatial coordinates". Then, the aforementioned node identifier is bound to the aforementioned multimodal feature data to generate multimodal feature nodes with unique identifiers, resulting in a multimodal feature node set. The binding can be set as a superscript. For example, A can represent the aforementioned multimodal feature data, B can represent the aforementioned node identifier, and the multimodal feature node with a unique identifier is A.B .

[0051] The second step involves encapsulating the feature vectors of each multimodal feature node in the aforementioned multimodal feature node set to generate structured feature nodes for the flying object, resulting in a set of structured feature nodes for the flying object. These structured feature nodes include: a feature vector, a timestamp, and the flying object's spatial coordinates. The timestamp represents the precise time the flying object was observed. The spatial coordinates represent the flying object's three-dimensional spatial position in a unified coordinate system, which can be represented by (x, y, z). The feature vector can be a 128-dimensional vector containing the flying object's features. These features may include shape features.

[0052] As an example, the aforementioned execution entity can extract the feature data, temporal information, and spatial information contained in each multimodal feature node in the aforementioned multimodal feature node set. First, the feature data is standardized to obtain a standardized feature vector; for example, all feature values ​​are normalized to the [0, 1] interval and arranged in descending order. Then, the temporal information is converted to a unified time format to obtain a timestamp, such as a Unix timestamp or ISO 8601 format. Next, the spatial coordinates are formatted into a unified three-dimensional coordinate representation to obtain three-dimensional spatial coordinates. Finally, the standardized feature vector, timestamp, and three-dimensional spatial coordinates are encapsulated into a structured data object to generate a structured feature node set for the flying object.

[0053] The third step is to generate temporal proximity, spatial proximity, and feature similarity values ​​based on the above set of structured feature nodes of flying objects.

[0054] As an example, the aforementioned execution entity can first select any two structured feature nodes from the aforementioned set of structured feature nodes of flying objects to calculate their temporal proximity, thereby generating a temporal proximity value. This temporal proximity value is inversely proportional to the time difference between the two structured feature nodes. The temporal proximity can be the time difference between the two structured feature nodes. Next, it selects any two structured feature nodes from the aforementioned set of structured feature nodes to calculate their spatial proximity, thereby generating a spatial proximity value. This spatial proximity value is inversely proportional to the spatial distance difference between the two structured feature nodes. The spatial proximity can be the distance difference between the two structured feature nodes. Finally, it selects any two structured feature nodes from the aforementioned set of structured feature nodes to calculate their feature similarity, thereby obtaining a feature similarity value. This feature similarity value is obtained using the cosine similarity calculation method.

[0055] The fourth step is to perform a weighted summation of the above-mentioned temporal proximity values, spatial proximity values, and feature similarity values ​​to obtain a comprehensive correlation strength value.

[0056] As an example, the aforementioned executing entity can use a weighted average algorithm to perform weighted and comprehensive processing on the aforementioned temporal proximity value, the aforementioned spatial proximity value, and the aforementioned feature similarity value to obtain a comprehensive association strength value.

[0057] Fifth, in response to the determination that the above-mentioned comprehensive association strength value exceeds a preset value threshold, flight trajectory association edges are established between the corresponding two structured feature nodes of the flying objects to generate a flight trajectory association edge set. The flight trajectory association edges in the above-mentioned flight trajectory association edge set include: flight trajectory association type, flight trajectory association strength, and flight trajectory directionality information.

[0058] Here, the aforementioned preset value threshold can be the maximum value of a pre-set value.

[0059] As an example, the aforementioned execution entity can connect the structured feature nodes of the aforementioned flying objects that satisfy the comprehensive association strength value exceeding a preset value threshold, and determine the edge connecting the two nodes as the flight trajectory association edge set.

[0060] The sixth step is to identify the structured feature nodes of the above-mentioned flying objects in the set of structured feature nodes as map nodes, thus obtaining the flying object map node set.

[0061] The seventh step is to construct a graph structure by combining the above-mentioned set of nodes in the flight object map and the set of edges associated with the flight trajectory, thereby obtaining an initial flight object feature map.

[0062] As an example, the aforementioned executing entity can connect the aforementioned set of flying object map nodes and the aforementioned set of flight trajectory associated edges to form a map structure of "nodes connecting edges connecting nodes", which serves as the initial flying object feature map.

[0063] The eighth step is to optimize the structure of the initial flying object feature map to generate a flying object perception feature map.

[0064] like Figure 5 As shown in the figure, a, b, c, d, e, f, and g represent different structured feature nodes of flying objects. The lines connecting the nodes represent the trajectory association edges, and the arrows indicate the trajectory direction. The trajectory association edges can be straight lines or curves. For example, a flying object flies from structured feature node a to structured feature node b.

[0065] As an example, the aforementioned executing entity can prune the initial flying object feature map to generate a pruned flying object feature map, which can then be used as a flying object perception feature map.

[0066] The content in steps one through eight above constitutes an inventive point of this disclosure, solving the technical problem of "low reliability of low-altitude defense schemes." Factors leading to interrupted flight paths and thus low reliability of low-altitude defense schemes often include: the difficulty in quantifying the correlation strength of flight objects in time, space, and feature dimensions due to reliance on human experience, resulting in isolated data points and the inability to form coherent flight paths. This can lead to interrupted flight paths, potentially causing the same flight object to be identified as two independent targets before and after the interruption, thus lowering the reliability of the low-altitude defense scheme. Solving these factors can improve the reliability of low-altitude defense schemes. To achieve this, firstly, in the first step, node identifiers are generated for each multimodal feature data in the aforementioned multimodal feature dataset to generate multimodal feature nodes with unique identifiers, resulting in a multimodal feature node set. These identifiers are generated based on a combination of time and spatial information. By combining timestamps and spatial coordinates to generate unique node identifiers, an identity is established for each data point, solving the problem of isolated and unrelated data points. The second step involves encapsulating the feature vectors of each multimodal feature node in the aforementioned multimodal feature node set to generate structured feature nodes for the flying object, thus obtaining a set of structured feature nodes for the flying object. This eliminates data format differences and improves data consistency and processability. The third step generates temporal proximity, spatial proximity, and feature similarity values ​​based on the aforementioned set of structured feature nodes for the flying object. The association strength between nodes is calculated from the three dimensions of time, space, and feature similarity, providing a comprehensive and objective association metric, replacing subjective human judgment, and improving the accuracy and consistency of association analysis. The fourth step involves weighted summation of the aforementioned temporal proximity, spatial proximity, and feature similarity values ​​to obtain a comprehensive association strength value. The fifth step, in response to the determination that the comprehensive association strength value exceeds a preset threshold, establishes a flight trajectory association edge between the corresponding two structured feature nodes for the flying object, thereby generating a flight trajectory association edge set. The flight trajectory association edges in the aforementioned flight trajectory association edge set include: flight trajectory association type, flight trajectory association strength, and flight trajectory directionality information. The process involves filtering weak associations using a preset threshold, retaining only significantly relevant node connections to avoid over-connection in the graph and improve association quality. The sixth step involves identifying the structured feature nodes of the aforementioned flight object structured feature node set as graph nodes, resulting in a flight object graph node set. The seventh step involves constructing a graph structure using the aforementioned flight object graph node set and the aforementioned flight trajectory association edge set, obtaining an initial flight object feature graph. This allows for the formation of coherent flight trajectories from the initial flight object feature graph, preventing trajectory interruptions and improving the reliability of the low-altitude defense scheme. The eighth step involves structural optimization of the initial flight object feature graph to generate a flight object perception feature graph.By performing optimization operations such as pruning, redundant connections and isolated nodes are removed, improving the quality and readability of the map and reducing computational complexity. Therefore, interruptions to the flight path are avoided, enhancing the reliability of low-altitude defense solutions.

[0067] Step 203: Extract the temporal node sequence from the above-mentioned flying object sensing feature map to generate a flying object temporal feature sequence.

[0068] In some embodiments, the aforementioned execution entity may extract the temporal node sequence from the aforementioned flying object sensing feature map to generate a flying object temporal feature sequence.

[0069] The aforementioned temporal feature sequence of flying objects represents the flight status changes of the same flying object at different times. For example, the aforementioned temporal feature sequence of flying objects could be the same flying object entering the detection range at 14:30 and continuously approaching the core area at 14:33. For example, the aforementioned temporal node sequence could be the time series from time t1 to time t2 corresponding to node A to node B, where node A (time t1, position p1), edge 1, and node B (time t2, position p2).

[0070] As an example, the aforementioned execution entity can extract temporally continuous nodes from the aforementioned object perception feature map to generate a temporally continuous feature node sequence, which serves as the object's temporal feature sequence. The temporally continuous feature nodes in this sequence are arranged sequentially according to their chronological order.

[0071] Step 204: Generate a threat level sequence based on the above-mentioned time-series characteristic sequence of flying objects.

[0072] In some embodiments, the aforementioned executing entity may generate a threat level sequence based on the aforementioned time-series characteristic sequence of the flying object.

[0073] The threat level sequence described above represents the threat level of an object at different trajectories over time. For example, the threat level sequence can include: no threat, low threat, medium threat, high threat, and severe threat. For instance, the threat level of an object entering the detection range at 14:30 is low threat. For example, for the flight path of an object from 14:30 to 14:35, the generated threat level sequence is: {[14:30-14:31: Low threat, 14:31-14:32: Medium threat, 14:32-14:33: High threat, 14:33-14:35: Severe threat]}.

[0074] As an example, the aforementioned execution entity can normalize the aforementioned temporal feature sequence of the flying object to generate a normalized temporal feature sequence of the flying object. A Support Vector Machine (SVM) is then used to perform binary classification on the normalized temporal feature sequence of the flying object, outputting probabilities greater than 0.8 for severe threat, [0.6, 0.8] for high threat, [0.3, 0.6] for medium threat, [0.1, 0.3] for low threat, and less than 0.1 for no threat, thus obtaining a threat level sequence frame by frame.

[0075] like Figure 6 As shown, the numbers 1-8 on the number line represent different distances from the core area, where 1 represents the closest position to the core area and 8 represents the farthest position. The core area can be the center of the ground. The broken line paths represent different flight trajectories of the flying object. The threat level of the trajectory between distances 1 and 2 is severe threat level; the threat level of the trajectory between distances 2 and 3 is high threat level; the threat level of the trajectory between distances 3 and 4 is medium threat level; the threat level of the trajectory between distances 4 and 5 is low threat level; and the threat level of the trajectory above distance 5 is no threat level. For example, the threat level of the flying object entering the detection range at 14:30 is low threat level.

[0076] Optionally, the aforementioned implementing entity can generate a threat level sequence based on the aforementioned time-series characteristic sequence of the flying object through the following steps:

[0077] The first step is to perform the following processing steps for each time series feature of the above-mentioned flight object time series feature sequence:

[0078] The first sub-step involves determining the behavioral anomaly degree of the aforementioned time-series characteristics of the flying object in order to generate a behavioral anomaly degree value.

[0079] Here, the aforementioned abnormal behavior level can characterize the degree of deviation from historical normal patterns. Historical normal patterns can be the regular patterns observed in historical monitoring data. For example, in the low-altitude area within 10 kilometers of an airport and below 500 meters in altitude, a historical normal pattern might be defined as: aircraft strictly following instrument approach procedures, speed > 150 knots, and a smooth trajectory. A few authorized mapping drones operate during the day, with flight speeds < 50 knots, and their activity range limited to specific areas.

[0080] As an example, the aforementioned implementing entity can first compare the time-series characteristics of the aircraft with a historical normal pattern database to generate a behavior anomaly score. The historical normal pattern database can be a database formed by normal patterns from multiple historical monitoring data, including the time-series characteristics of the aircraft and the corresponding behavior anomaly scores. For example, the historical normal pattern database could be: {Aircraft strictly follow instrument approach procedures, speed > 150 knots, smooth trajectory, behavior anomaly score: 5; A few authorized mapping drones operate during the day, flight speed < 50 knots, activity range limited to specific areas, behavior anomaly score: 6}.

[0081] The second sub-step involves determining the intent matching degree based on the aforementioned temporal characteristics of the flying object, in order to generate an intent matching degree value.

[0082] As an example, the aforementioned execution entity can perform feature matching processing on the node features in the temporal characteristics of the flying object with a preset intent pattern library to obtain a feature matching degree value. It can also perform sequence matching processing on the overall behavioral pattern of the flying object's temporal characteristics with a preset behavioral sequence template to obtain a sequence matching degree value. Furthermore, it can perform spatial relationship analysis processing on the spatial movement path of the flying object's temporal characteristics and the distribution of sensitive areas to obtain a spatial matching degree value. Finally, it can select the maximum value from the feature matching degree value, sequence matching degree value, and spatial matching degree value to obtain the intent matching degree value. The aforementioned preset intent pattern library can be a pre-established dataset containing typical behavioral characteristics of various threatening flying objects, used to compare with real-time flying object characteristics to determine their possible behavioral intentions. The aforementioned preset behavioral sequence template can be a time-series pattern describing the complete action process of a threatening flying object, including combinations of behavioral characteristics from multiple stages. The aforementioned spatial movement path can be the continuous trajectory of the flying object in three-dimensional space, including spatiotemporal information such as position, velocity, and direction. The aforementioned node features can characterize the complete state description of the flying object at a certain moment, and are a multi-dimensional feature vector. The multi-dimensional feature vector may include, but is not limited to, position coordinates and velocity components. Spatial relationship analysis can be a mathematical process that quantifies the degree of correlation between a spatial movement path and the distribution of sensitive areas by calculating the geometric, distance, topological, and motion relationships between the spatial movement path and the distribution of sensitive areas in space.

[0083] The third sub-step involves determining the environmental anomaly degree of the aforementioned time-series characteristics of the flying objects to generate an environmental anomaly degree value.

[0084] As an example, the aforementioned executing entity can first obtain environmental parameters of the flight object's temporal characteristics, including time information and terrain information. Then, it compares the occurrence time of the flight object's temporal characteristics with a preset activity schedule to obtain a time anomaly value. The preset activity schedule can be from 9:00 AM to 11:00 AM and from 1:00 PM to 8:00 PM. Next, it compares the flight object's temporal movement pattern with feasible movement patterns under the current terrain conditions to obtain a terrain anomaly value. Feasible movement patterns under the current terrain conditions can be no-fly terrain conditions. Finally, it weights and combines the aforementioned time anomaly values ​​and terrain anomaly values ​​to obtain an environmental anomaly value.

[0085] The fourth sub-step involves determining the threat score based on the aforementioned behavioral anomaly score, intent matching score, and environmental anomaly score to generate a threat score value.

[0086] As an example, the aforementioned executing entity can determine a threat score by using a preset weight set to evaluate the aforementioned behavioral anomaly value, intent matching value, and environmental anomaly value. For instance, the preset weight set may include: weight 1, weight 2, and weight 3, where weight 1 + weight 2 + weight 3 = 1, and the threat score = behavioral anomaly value × weight 1 + intent matching value × weight 2 + environmental anomaly value × weight 3.

[0087] The fifth sub-step involves establishing a mapping table between preset threat scores and threat levels. This mapping table includes: no threat level and corresponding score, low threat level and corresponding score, medium threat level and corresponding score, high threat level and corresponding score, and severe threat level and corresponding score. For example, the mapping table could be {no threat level: (0, 1), low threat level: (1, 3), medium threat level: (3, 5), high threat level: (5, 7), severe threat level: (7, 10)}. The preset threat scores can be pre-defined scores representing the degree of threat posed by the flying object. For example, 9 points represents a severe threat.

[0088] The second step is to compare the obtained threat score set with the above mapping table to obtain the corresponding threat level sequence.

[0089] As an example, the aforementioned execution entity can match each threat score value in the obtained threat score set with the aforementioned mapping table to generate the corresponding threat level, thus obtaining the corresponding threat level sequence. For example, a threat score value of 4 corresponds to a medium threat level. The matching process can involve comparing each threat score value with the aforementioned mapping table.

[0090] Step 205: Parameterize the above threat level sequence to generate a parameterized flying object scenario model.

[0091] In some embodiments, the aforementioned execution entity may perform scenario parameterization on the aforementioned threat level sequence to generate a parameterized flying object scenario model.

[0092] As an example, the aforementioned execution entity can first extract the characteristic parameters of the flying objects from the threat level sequence. For example, the characteristic parameters could be the current threat level, flying object type, and motion state. Then, it can obtain the current environmental condition parameters, such as time, weather, visibility, and wind speed. Next, it can query available resource status parameters, such as sensor status, number of response units, and communication quality. Finally, it can combine the extracted flying object characteristic parameters, environmental condition parameters, and resource status parameters into a structured parameter set to generate a parameterized flying object scenario model. The parameterized flying object scenario model can be (S, A, T), containing a quantitative description of multiple dimensions, including flying object parameters, environmental condition parameters, and resource status parameters. Here, S represents the state space, containing the states of all parameters, such as S = {s_flying object, s_environment, s_resource, s_time}. A represents the action space, such as A = {acceleration, turning}. T represents the transition function, describing how the state changes under the action. For example, a parameterized flight object scenario model can be a dynamic adversarial environment model, which simplifies the real-world scenario into a deductive digital twin model.

[0093] Optionally, the aforementioned implementing entity can perform scenario parameterization on the aforementioned threat level sequence through the following steps to generate a parameterized flying object scenario model:

[0094] The first step is to parameterize the environmental conditions in which the above threat level sequence occurs to generate an environmental parameter set. This environmental parameter set includes: wind speed parameters, terrain parameters, and time parameters.

[0095] As an example, the aforementioned implementing entity can determine environmental parameters of the environment in which the aforementioned threat level sequence occurs to generate a set of environmental parameters. For example, the aforementioned environmental parameters can be terrain parameters (e.g., mountains, plains). For example, from the actual physical environment in which the threat level sequence occurs, various environmental elements affecting low-altitude defense operations can be systematically extracted, quantified, and organized.

[0096] The second step is to parameterize the response resources of the above threat level sequence to generate a resource parameter set, which includes: sensor parameters and communication parameters.

[0097] As an example, the aforementioned executing entity can determine resource parameters for the response resources of the aforementioned threat level sequence to generate a resource parameter set. For example, the aforementioned resource parameters can be communication status parameters (e.g., network quality, bandwidth limitations).

[0098] The third step is to integrate the above environmental parameter set and the above resource parameter set into a model to generate a parameterized flying object scene model.

[0099] As an example, the aforementioned execution entity can populate the aforementioned environmental parameter set and resource parameter set into a preset data structure to generate a parameterized flying object scene model. For example, the parameterized flying object scene model can be {"Environmental Parameters": {"Time Parameters": ["Timestamp": "2025-12-29 10:30:20"], "Geographical Parameters": ["Terrain": "City"]}, "Resource Parameters": {"Sensor Resources": ["Type": "Radar", "Status": "Normal"]}. The aforementioned preset data structure can be an empty two-dimensional array.

[0100] Step 206: Based on the above parameterized flying object scenario model, determine the target low-altitude defense execution plan.

[0101] In some embodiments, the aforementioned executing entity may determine a target low-altitude defense execution plan based on the aforementioned parameterized flying object scenario model. Here, the aforementioned target low-altitude defense execution plan is an execution plan for defending against the aforementioned low-altitude flying objects.

[0102] As an example, the aforementioned implementing entity can use a lookup table to query the threat level of the parameterized flying object scenario model to obtain the target low-altitude defense implementation plan. For instance, the handling plan corresponding to a high threat level is microwave jamming.

[0103] Optionally, the aforementioned implementing entity can determine the target low-altitude defense implementation plan based on the above-mentioned parameterized flying object scenario model through the following steps:

[0104] The first step is to read the temporal feature groups corresponding to the preset threat level from the above parameterized flying object scenario model.

[0105] Here, the preset threat level can be a pre-defined "low threat level, medium threat level, high threat level, or severe threat level".

[0106] The second step is to perform level mapping processing on the above time series feature groups to generate response level groups.

[0107] As an example, the aforementioned implementing entity can determine the time-series features representing low threat levels from the aforementioned time-series feature group as the monitoring response level. The time-series features representing medium threat levels from the aforementioned time-series feature group can be determined as the warning response level. The time-series features representing high threat levels from the aforementioned time-series feature group can be determined as the interception response level. The time-series features representing severe threat levels from the aforementioned time-series feature group can be determined as the handling response level. Finally, the aforementioned monitoring response level, warning response level, interception response level, and handling response level are determined as a response level group. The aforementioned monitoring response level can be a real-time monitoring response level, the aforementioned warning response level can be a warning-issuing response level, the aforementioned interception response level can be an interception-issuing response level, and the aforementioned handling response level can be a handling-issuing response level.

[0108] The third step is to select candidate action plans corresponding to the above response level groups from the predefined action plan library to obtain a set of candidate action plans.

[0109] Here, the aforementioned predefined action plan library can be a pre-set database containing action plans. For example, the aforementioned predefined action plan library can be {monitoring level action: continuously track, identify and record the target; warning level action: warn the target through non-contact means, forcing it to change course or leave; interception level action: use interception means (such as netting or jamming) to force the target to lose control or land; disposal level action: use authorized force to eliminate the threat}.

[0110] As an example, the aforementioned execution entity can select the corresponding action plan from the predefined action plan library for each response level in the response level group to generate candidate action plans and obtain a set of candidate action plans.

[0111] The fourth step is to perform simulation analysis on the above candidate action plan set to generate a simulation analysis result set.

[0112] As an example, the aforementioned executing entity can decompose each candidate action plan in the aforementioned candidate action plan set into execution steps to generate a step sequence. Then, it performs a single-step simulation of the step sequence to obtain a single-step execution result set. The steps in the single-step execution result set are then ordered sequentially to generate the simulation process for a single candidate action plan. The resource consumption and simulation effects of the simulation process are determined to generate simulation resource consumption and simulation effects. Finally, the simulation resource consumption and simulation effects are determined as the simulation results.

[0113] As another example, the aforementioned implementing entity can use MATLAB or Simulink to simulate and extrapolate the aforementioned set of candidate action plans to generate a set of simulation results.

[0114] The fifth step is to perform multi-objective optimization processing on the above simulation results set to determine the target low-altitude defense execution plan.

[0115] As an example, the aforementioned implementing entity can use a multi-objective optimization algorithm to perform multi-objective optimization processing on the simulation result set to determine the target low-altitude defense execution plan. The aforementioned multi-objective optimization algorithm can be the Top-Level Optimal Solution (TOPSIS). For example, TOPSIS sorts each simulation result by calculating the relative distance between it and the target's best solution (where all targets are optimal) and worst solution (where all targets are worst), selecting the solution closest to the target's best and furthest from the target's worst.

[0116] Step 207: In response to determining that the execution deviation value corresponding to the above-mentioned target low-altitude defense execution plan exceeds a preset deviation threshold, the above-mentioned target low-altitude defense execution plan is dynamically adjusted.

[0117] In some embodiments, the execution entity may dynamically adjust the target low-altitude defense execution scheme in response to determining that the execution deviation value corresponding to the target low-altitude defense execution scheme exceeds a preset deviation threshold.

[0118] Here, the aforementioned execution deviation value characterizes the degree of difference between the actual execution state and the expected state. The aforementioned preset deviation threshold can be a pre-set deviation critical value. The specific value of the preset deviation threshold is not limited here.

[0119] As an example, the aforementioned implementing entity may first, in response to determining that the execution deviation value corresponding to the aforementioned low-altitude defense execution plan represents an environmental change deviation, activate the backup sensor for the aforementioned low-altitude defense execution plan. In response to determining that the execution deviation value corresponding to the aforementioned low-altitude defense execution plan represents a resource status deviation, activate the backup communication link for the aforementioned low-altitude defense execution plan. The backup sensor may be a backup radar sensor. The aforementioned environmental change deviation may represent a decrease in visibility. The aforementioned resource status deviation may represent a communication link interruption. The backup communication link may be a backup communication system capable of taking over or supplementing the function of the main communication link when the main communication link fails or its performance is severely degraded.

[0120] Optionally, the aforementioned implementing entity may dynamically adjust the aforementioned low-altitude air defense implementation plan in response to determining that the execution deviation value corresponding to the aforementioned target low-altitude defense implementation plan exceeds a preset deviation threshold through the following steps:

[0121] The first step is to break down the above-mentioned low-altitude defense implementation plan into steps to generate a sequence of action steps.

[0122] As an example, the aforementioned execution entity can break down the target low-altitude defense execution plan into multiple individual steps. Then, the execution order of these individual steps is determined, resulting in a step sequence. This breakdown can be a combination of logical decomposition and sequential determination.

[0123] The second step is to determine the time window for each action step in the above action step sequence to generate an action step time window set.

[0124] Here, the time window described above can be a specific time interval allocated to a single step. For example, the time window can include: start time, end time, and duration. The start time represents the relative point in time when the step begins. The end time represents the relative point in time when the step completes. The duration represents the maximum time the step is allowed to execute.

[0125] The third step is to allocate resources for each action step in the above action step sequence to generate action step resource information, resulting in an action step resource information set. The action step resource information in the above action step resource information set includes: resource type and resource quantity.

[0126] Here, the resource type mentioned above can be the category of the required resource. The resource quantity mentioned above can be the quantity or capacity required for each type of resource. For example, two interceptor drones (resource quantity) (resource category).

[0127] As an example, the executing entity can first determine the resource requirements for each action step in the sequence of action steps to generate an action step resource requirement set. Then, it can allocate corresponding resources to this action step resource requirement set to obtain an action step resource information set. These resource requirements can be for different resource types.

[0128] The fourth step involves encapsulating the above sequence of action steps, the set of time windows for the above action steps, and the set of resource information for the above action steps to generate an executable low-altitude defense action plan.

[0129] As an example, the aforementioned implementing entity can integrate the above-mentioned action step sequence, action step time window set, and action step resource information set into an executable low-altitude defense action plan. This integration can also be a combination.

[0130] The fifth step is to obtain information on the actual low-altitude defense execution status of the above-mentioned executable low-altitude defense action plan during its execution.

[0131] Here, the aforementioned actual low-altitude defense execution status information can be information collected through sensors and other means during the implementation of the plan. For example, this actual low-altitude defense execution status information may include, but is not limited to: the actual start and end times of each step, actual resource usage, and environmental changes during execution. Sensors may include, but are not limited to: radar, photoelectric cameras.

[0132] The sixth step is to compare the actual low-altitude defense execution status information with the preset execution status to obtain the execution deviation value.

[0133] Here, the above-mentioned preset execution state can be "the plan is to execute at 10:30, 70 meters away from the center of the ground along the x-axis".

[0134] As an example, the aforementioned implementing entity can compare the actual low-altitude defense execution status information with the preset execution status to obtain the time deviation value. It can also compare the actual low-altitude defense execution status information with the preset execution status to obtain the position deviation value. The average of the time deviation value and the position deviation value is determined as the execution deviation value. For example, the actual low-altitude defense execution status information is "executed at 10:30:12 at a distance of 85 meters from the center of the ground along the x-axis". The time deviation value is 12 seconds. The position deviation value is 15 meters. The execution deviation value is (12 / 60 + 15 / 70) / 2 ≈ 0.2.

[0135] Step 7: In response to determining that the aforementioned execution deviation value exceeds a preset deviation threshold, a preset multi-dimensional score is performed on the aforementioned low-altitude defense execution plan to generate a multi-dimensional plan score set. The multi-dimensional plan scores in the aforementioned multi-dimensional plan score set include scores for different plan steps.

[0136] Here, the aforementioned preset multiple dimensions can be pre-defined dimensions such as time efficiency, resource efficiency, and risk control. The multi-dimensional solution score could be {Time Efficiency: 50 points, Resource Efficiency: 80 points, Risk Control: 90 points}.

[0137] As an example, the aforementioned implementing entity can score the execution plan for low-altitude air defense based on its time efficiency to generate a time efficiency score. For instance, if the preset time threshold is 50 seconds for execution to complete, and the execution plan completes within 100 seconds, the time efficiency score is (100-50) / 100 = 50 points. The execution plan can also be scored on its resource efficiency to generate a resource efficiency score. For example, reasonable equipment usage scores 80 points. Finally, the execution plan can be scored on its risk control to generate a risk control score. These time efficiency scores, resource efficiency scores, and risk control scores are then combined to form a multi-dimensional plan score set.

[0138] The eighth step involves adjusting the parameters of the steps corresponding to the multi-dimensional schemes whose scores are lower than the preset scoring threshold in the above multi-dimensional scheme scoring set, in order to generate the target low-altitude defense execution scheme.

[0139] As an example, the aforementioned implementing entity can adjust the interception distance and the number of drones for the steps corresponding to multi-dimensional schemes with scores below a preset scoring threshold in the multi-dimensional scheme scoring set, in order to generate a target low-altitude defense execution scheme. The aforementioned interception distance adjustment can be from 100 meters to 80 meters. The aforementioned drone number adjustment can be from 1 to 2.

[0140] The content of steps one through eight above constitutes an inventive point of this disclosure, solving the technical problem of "difficulty in timely adjustment of low-altitude defense schemes." Factors leading to the difficulty in timely adjustment of low-altitude defense schemes often include: fixed target low-altitude defense execution schemes cannot adapt to dynamically changing threat behaviors and environmental conditions; the execution schemes exhibit poor robustness and insufficient dynamic adaptability when facing complex and ever-changing low-altitude threats, making it difficult to adjust the low-altitude defense scheme in a timely manner when emergencies occur. Solving these factors can achieve the effect of timely adjustment of low-altitude defense schemes. To achieve this effect, firstly, the target low-altitude defense execution scheme is decomposed into steps to generate an action step sequence. This facilitates subsequent steps. Secondly, a time window is determined for each action step in the action step sequence to generate an action step time window set. A precise time interval is allocated to each step to avoid timing conflicts. Thirdly, resources are allocated for each action step in the action step sequence to generate action step resource information, resulting in an action step resource information set. The action step resource information in the action step resource information set includes: resource type and resource quantity. The fourth step involves encapsulating the above action step sequence, action step time window set, and above action step resource information set to generate an executable low-altitude defense action plan. The fifth step involves obtaining the actual low-altitude defense execution status information during the execution of the executable low-altitude defense action plan. The sixth step involves comparing the actual low-altitude defense execution status information with the preset execution status to obtain the execution deviation value. This allows for timely identification of execution anomalies. The seventh step, in response to the determination that the execution deviation value exceeds a preset deviation threshold, involves performing a preset multi-dimensional score on the target low-altitude defense execution plan to generate a multi-dimensional plan score set. This multi-dimensional plan score set includes scores for different plan steps. Adapting to dynamically changing threat behaviors and environmental conditions through multi-dimensional scoring helps improve the robustness and dynamic adaptability of the execution plan when facing complex and ever-changing low-altitude threats. The eighth step involves adjusting the parameters of plan steps whose scores in the multi-dimensional plan score set are lower than the preset score threshold to generate the target low-altitude defense execution plan. Adjusting parameters can address the challenge of timely adjustments to low-altitude defense strategies in the event of unforeseen circumstances. Therefore, it can adapt to dynamically changing threat behaviors and environmental conditions, contributing to improved robustness and dynamic adaptability of the execution plan when facing complex and ever-changing low-altitude threats.

[0141] Further reference Figure 3 As an implementation of the methods shown in the above figures, this disclosure provides some embodiments of a low-altitude defense scheme adjustment device based on multimodal feature fusion. These device embodiments are similar to... Figure 2Corresponding to the method embodiments shown, this low-altitude defense scheme adjustment device based on multimodal feature fusion can be specifically applied to various electronic devices.

[0142] like Figure 3 As shown, a low-altitude defense scheme adjustment device 300 based on multimodal feature fusion in some embodiments includes: a conversion unit 301, an encoding unit 302, an extraction unit 303, a generation unit 304, a parameterization unit 305, a determination unit 306, and a dynamic adjustment unit 307. The conversion unit 301 is configured to convert a low-altitude related dataset for a low-altitude target detection scenario into a multimodal feature dataset. The low-altitude target detection scenario represents a scenario where undetected low-altitude flying objects exist in a low-altitude area within a preset range. The low-altitude related dataset is a dataset related to low-altitude flying objects collected in real time by multiple pre-deployed sensors. The encoding unit 302 is configured to perform associative encoding on the multimodal feature dataset to generate a flying object perception feature map. Each node of the flying object perception feature map represents flying object feature data at different times, and each edge of the perception feature map represents the flying object's flight trajectory. The extraction unit 303 is configured to extract a temporal node sequence from the flying object perception feature map to generate a flying object temporal feature sequence. The aforementioned temporal feature sequence of the flying object represents the flight state changes of the same flying object at different times; the generation unit 304 is configured to generate a threat level sequence based on the aforementioned temporal feature sequence of the flying object, wherein the aforementioned threat level sequence represents the threat level of the flying object at different trajectories as time changes; the parameterization unit 305 is configured to perform scene parameterization on the aforementioned threat level sequence to generate a parameterized flying object scene model; the determination unit 306 is configured to determine a target low-altitude defense execution plan based on the aforementioned parameterized flying object scene model, wherein the aforementioned target low-altitude defense execution plan is an execution plan for defending against the aforementioned low-altitude flying object; the dynamic adjustment unit 307 is configured to dynamically adjust the aforementioned target low-altitude defense execution plan in response to the determination that the execution deviation value corresponding to the aforementioned target low-altitude defense execution plan exceeds a preset deviation threshold.

[0143] It is understandable that the units described in the low-altitude defense scheme adjustment device 300 based on multimodal feature fusion are related to the reference... Figure 2 The steps in the described method correspond accordingly. Therefore, the operations, features, and beneficial effects described above for the method are also applicable to the low-altitude defense scheme adjustment device 300 based on multimodal feature fusion and the units contained therein, and will not be repeated here.

[0144] The following is for reference. Figure 4 It shows a schematic diagram of the structure of an electronic device 400 (e.g., a computing device) suitable for implementing some embodiments of the present disclosure. Figure 4 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments of this disclosure.

[0145] like Figure 4 As shown, the electronic device 400 may include a processing unit 401 (e.g., a central processing unit, a graphics processor, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 402 or a program loaded from a storage device 408 into a random access memory (RAM) 403. The RAM 403 also stores various programs and data required for the operation of the electronic device 400. The processing unit 401, ROM 402, and RAM 403 are interconnected via a bus 404. An input / output (I / O) interface 405 is also connected to the bus 404.

[0146] Typically, the following devices can be connected to I / O interface 405: input devices 406 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 407 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 408 including, for example, magnetic tapes, hard disks, etc.; and communication devices 409. Communication device 409 allows electronic device 400 to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 4 An electronic device 400 with various devices is shown; however, it should be understood that it is not required to implement or possess all of the devices shown. More or fewer devices may be implemented or possessed alternatively. Figure 4 Each box shown can represent a device or multiple devices as needed.

[0147] In particular, according to some embodiments of this disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, some embodiments of this disclosure include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication device 409, or installed from storage device 408, or installed from ROM 402. When the computer program is executed by processing device 401, it performs the functions defined above in the methods of some embodiments of this disclosure.

[0148] It should be noted that, in some embodiments of this disclosure, the computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium may be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In some embodiments of this disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In some embodiments of this disclosure, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wires, optical fibers, RF (radio frequency), etc., or any suitable combination thereof.

[0149] In some implementations, clients and servers can communicate using any currently known or future-developed network protocol such as HTTP (Hypertext Transfer Protocol) and can interconnect with digital data communication (e.g., communication networks) of any form or medium. Examples of communication networks include local area networks (“LANs”), wide area networks (“WANs”), the Internet (e.g., the Internet of Things), and peer-to-peer networks (e.g., ad hoc peer-to-peer networks), as well as any currently known or future-developed networks.

[0150] The aforementioned computer-readable medium may be included in the aforementioned electronic device; or it may exist independently and not assembled into the electronic device. The aforementioned computer-readable medium carries one or more programs, which, when executed by the electronic device, cause the electronic device to: convert the collected low-altitude related dataset for low-altitude target detection scenarios into a multimodal feature dataset, wherein the aforementioned low-altitude target detection scenario represents a scenario where undetected low-altitude flying objects exist in a low-altitude area within a preset range, and the aforementioned low-altitude related dataset is a dataset related to low-altitude flying objects collected in real time by multiple pre-deployed sensors; perform association encoding on the aforementioned multimodal feature dataset to generate a flying object perception feature map, wherein each node of the aforementioned flying object perception feature map represents flying object feature data at different times, and each edge of the aforementioned perception feature map represents the flight trajectory of the flying object; and perform association encoding on the aforementioned flying object perception feature map. The system extracts temporal node sequences from the target image to generate a temporal feature sequence of the flying object, where the temporal feature sequence represents the flight state changes of the same flying object at different times. Based on the temporal feature sequence, a threat level sequence is generated, where the threat level sequence represents the threat level of the flying object at different trajectories over time. The threat level sequence is then parameterized to generate a parameterized flying object scenario model. Based on the parameterized flying object scenario model, a target low-altitude defense execution plan is determined, where the target low-altitude defense execution plan is the execution plan for defending against the low-altitude flying object. In response to the determination that the execution deviation value corresponding to the target low-altitude defense execution plan exceeds a preset deviation threshold, the target low-altitude defense execution plan is dynamically adjusted.

[0151] Computer program code for performing operations of some embodiments of this disclosure can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0152] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0153] The functions described above in this document can be performed at least in part by one or more hardware logic components. For example, exemplary types of hardware logic components that can be used, without limitation, include: field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), system-on-a-chip (SoCs), complex programmable logic devices (CPLDs), and so on.

[0154] The above description is merely a selection of preferred embodiments of this disclosure and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in the embodiments of this disclosure is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-described inventive concept. For example, technical solutions formed by substituting the above-described features with (but not limited to) technical features with similar functions disclosed in the embodiments of this disclosure.

Claims

1. A method for adjusting a low-altitude defense scheme based on multimodal feature fusion, characterized in that, include: The collected low-altitude related datasets for low-altitude target detection scenarios are transformed into multimodal feature datasets. The low-altitude target detection scenario represents a scenario in which there are undetected low-altitude flying objects in a low-altitude area within a preset range. The low-altitude related datasets are datasets related to low-altitude flying objects collected in real time by multiple pre-deployed sensors. The multimodal feature dataset is correlated and encoded to generate an object perception feature map, wherein each node of the object perception feature map represents object feature data at different times, and each edge of the perception feature map represents the flight trajectory of the object. The temporal node sequence of the perceived feature map of the flying object is extracted to generate a temporal feature sequence of the flying object, wherein the temporal feature sequence of the flying object represents the flight state changes of the same flying object at different times; Based on the time-series characteristic sequence of the flying object, a threat level sequence is generated, wherein the threat level sequence represents the threat level of the flying object at different trajectories that change over time; The threat level sequence is parameterized to generate a parameterized flying object scenario model; Based on the parameterized flying object scenario model, a target low-altitude defense execution plan is determined, wherein the target low-altitude defense execution plan is an execution plan for defending against the low-altitude flying object; In response to determining that the execution deviation value corresponding to the target low-altitude defense execution plan exceeds a preset deviation threshold, the target low-altitude defense execution plan is dynamically adjusted.

2. The method according to claim 1, characterized in that, The process of converting the collected low-altitude related datasets for low-altitude target detection scenarios into a multimodal feature dataset includes: The low-altitude related dataset is timestamped to generate an aligned low-altitude related dataset, wherein each aligned low-altitude related data in the aligned low-altitude related dataset has a unified timestamp sequence; The aligned low-altitude correlation dataset is subjected to spatial coordinate transformation to generate a transformed low-altitude correlation dataset, wherein each transformed low-altitude correlation data in the transformed low-altitude correlation dataset is mapped to the same spatial coordinate system; The aligned low-altitude correlation dataset and the transformed low-altitude correlation dataset are subjected to spatiotemporal registration and fusion processing to generate a spatiotemporal aligned dataset. Each spatiotemporal aligned data in the spatiotemporal aligned dataset includes: position data, velocity data and heading data of the flying object under the same spatiotemporal information. Multimodal feature extraction processing is performed on the spatiotemporal aligned dataset to generate a multimodal feature dataset.

3. The method according to claim 1, characterized in that, The step of generating a threat level sequence based on the temporal feature sequence of the flying object includes: For each time-series feature of the aircraft in the time-series feature sequence, the following processing steps are performed: The behavioral anomaly degree is determined based on the temporal characteristics of the flying object to generate a behavioral anomaly degree value; The intent matching degree is determined based on the temporal characteristics of the flying object to generate an intent matching degree value; The environmental anomaly degree is determined based on the temporal characteristics of the flying object to generate an environmental anomaly degree value; Threat scores are determined based on the behavioral anomaly score, the intent matching score, and the environmental anomaly score to generate a threat score value. Establish a mapping table between preset threat scores and threat levels, wherein the mapping table includes: no threat level and corresponding score, low threat level and corresponding score, medium threat level and corresponding score, high threat level and corresponding score, and severe threat level and corresponding score; The obtained threat score set is compared with the mapping table to obtain the corresponding threat level sequence.

4. The method according to claim 1, characterized in that, The step of parameterizing the threat level sequence to generate a parameterized flying object scene model includes: The environment in which the threat level sequence occurs is parameterized to generate an environmental parameter set, wherein the environmental parameter set includes: wind speed parameter, terrain parameter and time parameter; The response resources for the threat level sequence are parameterized to generate a resource parameter set, wherein the resource parameter set includes: sensor parameters and communication parameters; The environmental parameter set and the resource parameter set are integrated into a model to generate a parameterized flying object scene model.

5. The method according to claim 1, characterized in that, The determination of the target low-altitude defense execution plan based on the parameterized flight object scenario model includes: Read the temporal feature group corresponding to the preset threat level from the parameterized flying object scene model; The time-series feature groups are subjected to level mapping processing to generate response level groups; Select candidate action plans corresponding to the response level group from the predefined action plan library to obtain a candidate action plan set; The candidate action plan set is subjected to simulation and deduction processing to generate a simulation and deduction result set; The simulation results set is subjected to multi-objective optimization processing to determine the target low-altitude defense execution plan.

6. A low-altitude defense scheme adjustment device based on multimodal feature fusion, characterized in that, include: The conversion unit is configured to convert the collected low-altitude related dataset for low-altitude target detection scenarios into a multimodal feature dataset. The low-altitude target detection scenario represents a scenario in which there are undetected low-altitude flying objects in a low-altitude area within a preset range. The low-altitude related dataset is a dataset related to low-altitude flying objects collected in real time by a variety of pre-deployed sensors. The encoding unit is configured to perform associative encoding on the multimodal feature dataset to generate an object perception feature map, wherein each node of the object perception feature map represents object feature data at different times, and each edge of the perception feature map represents the flight trajectory of the object. The extraction unit is configured to extract the temporal node sequence from the perceived feature map of the flying object to generate a temporal feature sequence of the flying object, wherein the temporal feature sequence of the flying object represents the flight state changes of the same flying object at different times; The generation unit is configured to generate a threat level sequence based on the temporal feature sequence of the flying object, wherein the threat level sequence represents the threat level of the flying object at different trajectories that varies over time; The parameterization unit is configured to perform scene parameterization on the threat level sequence to generate a parameterized flying object scene model; The determining unit is configured to determine a target low-altitude defense execution plan based on the parameterized flying object scenario model, wherein the target low-altitude defense execution plan is an execution plan for defending against the low-altitude flying object; The dynamic adjustment unit is configured to dynamically adjust the target low-altitude defense execution scheme in response to determining that the execution deviation value corresponding to the target low-altitude defense execution scheme exceeds a preset deviation threshold.

7. An electronic device, characterized in that, include: One or more processors; A storage device on which one or more programs are stored; When the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in any one of claims 1 to 5.

8. A computer-readable medium, characterized in that, It stores a computer program thereon, wherein the computer program, when executed by a processor, implements the method as described in any one of claims 1 to 5.

Images