Dynamic Limit Directional Positioning Method and System for Multi-Platform Joint Search and Rescue at Sea

Abstract

The embodiment of this application belongs to the technical field of machine learning, and discloses a dynamic limit directional positioning method and system for multi-platform joint search and rescue at sea. The method includes: mining features from the source direction-finding interaction data to obtain a linkage direction-finding interaction vector; obtaining a target multi-platform collaborative direction-finding code; determining a target linkage direction-finding interaction vector through the dynamic direction-finding quantization result of the direction-finding information of the first reporting device in the linkage direction-finding interaction vector and the target multi-platform collaborative direction-finding code; generating a spatio-temporal directional positioning label corresponding to the regional joint search and rescue direction-finding record according to the target linkage direction-finding interaction vector. In this way, the position and state of the target reporting device can be accurately determined in the spatio-temporal dimension, thereby greatly improving the search and rescue efficiency in the regional joint search and rescue scenario, enhancing the accuracy and credibility of target positioning, and reducing problems such as search and rescue delays caused by inaccurate positioning.

Description

Technical Field

[0001] The embodiments of the present application belong to the field of machine learning technology, and specifically relate to a dynamic limited directional positioning method and system applied to multi-platform joint search and rescue at sea. Background Art

[0002] In maritime search and rescue scenarios, accurately determining the location and status of target notification devices is crucial. Traditional positioning technologies often rely solely on direction-finding information from a single platform or fail to fully consider the real-time status of target notification devices. This results in low positioning accuracy and poor reliability in complex environments. For example, during maritime search and rescue, relying solely on the direction-finding equipment of a single vessel to locate the signal source of a wrecked vessel is susceptible to interference from factors such as the ocean environment and equipment errors, resulting in inaccurate positioning results and delaying search and rescue efforts. Furthermore, without combining direction-finding information with the real-time status of the target device, the accuracy and reliability of positioning cannot be fully assessed. Summary of the Invention

[0003] The embodiments of the present application provide a dynamic limited directional positioning method and system for multi-platform joint search and rescue at sea, which can solve or partially solve the technical problems involved in the above-mentioned background technology.

[0004] An embodiment of the present application provides a dynamic limited directional positioning method for multi-platform joint search and rescue at sea, which is applied to a dynamic limited directional positioning system. The method includes: performing feature mining on source direction finding interaction data to obtain a linkage direction finding interaction vector; the source direction finding interaction data is used to describe a regional joint search and rescue direction finding record in which a target notification device exists, and the source direction finding interaction data includes first notification device direction finding information representing the target notification device; the linkage direction finding interaction vector includes a dynamic direction finding quantization result of the first notification device direction finding information; obtaining a target multi-platform collaborative direction finding code; the target multi-platform collaborative direction finding code is obtained by integrating a state monitoring description vector of first source state monitoring data on the basis of the dynamic direction finding quantization result of the first notification device direction finding information, the first source state monitoring data representing the real-time monitoring state of the target notification device; determining a target linkage direction finding interaction vector through the dynamic direction finding quantization result of the first notification device direction finding information in the linkage direction finding interaction vector and the target multi-platform collaborative direction finding code; and generating a spatiotemporal directional positioning tag corresponding to the regional joint search and rescue direction finding record based on the target linkage direction finding interaction vector.

[0005] An embodiment of the present application provides a dynamic limit directional positioning system, comprising at least one processor and a memory; the memory stores computer-executable instructions; the at least one processor executes the computer-executable instructions stored in the memory, so that the at least one processor executes the above-mentioned method.

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

[0007] In the embodiment of the present application, first, by performing feature mining on the source direction-finding interaction data, a linkage direction-finding interaction vector is obtained, and the original direction-finding information is converted into a dynamic direction-finding quantization result, which can effectively integrate the direction-finding information of multiple platforms and make it more valuable for processing; second, obtaining the target multi-platform collaborative direction-finding coding integrates the direction-finding result and the real-time status monitoring data of the target reporting device, which can make the positioning information more comprehensive; then, determining the target linkage direction-finding interaction vector and further integrating multi-source information can improve the accuracy and reliability of the information. Finally, generating a spatio-temporal orientation positioning label based on the target linkage direction-finding interaction vector can accurately determine the position and status of the target reporting device in the spatio-temporal dimension, which greatly improves the search and rescue efficiency in the regional joint search and rescue scenario, enhances the accuracy and credibility of target positioning, and reduces problems such as search and rescue delays caused by inaccurate positioning. Brief Description of the Drawings

[0008] Figure 1 It is a flowchart of a dynamic limit orientation positioning method applied to multi-platform joint search and rescue at sea provided by an embodiment of the present application.

[0009] Figure 2 It is a schematic structural diagram of a dynamic limit orientation positioning system provided by an embodiment of the present application. Detailed Embodiments

[0010] Figure 1 A dynamic limit orientation positioning method applied to multi-platform joint search and rescue at sea is shown, which is applied to a dynamic limit orientation positioning system, and the method includes the following steps 110-step 140.

[0011] Step 110: Perform feature mining on the source direction-finding interaction data to obtain a linkage direction-finding interaction vector.

[0012] Among them, the source direction-finding interaction data is used to describe the regional joint search and rescue direction-finding record where the target reporting device exists, and the source direction-finding interaction data includes the first reporting device direction-finding information representing the target reporting device; the dynamic direction-finding quantization result of the first reporting device direction-finding information is included in the linkage direction-finding interaction vector.

[0013] For example, the linked direction finding interaction vector is a vector obtained by performing feature mining on the source direction finding interaction data. The source direction finding interaction data describes the joint search and rescue direction finding record of the area where the target notification device exists, which includes the direction finding information of the first notification device that represents the target notification device, while the linked direction finding interaction vector includes elements such as the dynamic direction finding quantization result of the first notification device direction finding information. In a maritime search and rescue scenario, for example, there are three platforms performing direction finding on the target notification device. The azimuth angle obtained by platform A is 30° and the distance is 10 nautical miles; the azimuth angle obtained by platform B is 45° and the distance is 12 nautical miles; the azimuth angle obtained by platform C is 60° and the distance is 15 nautical miles. The vector obtained by performing feature mining on these direction finding information, which includes elements such as the quantized results of these azimuth and distance information, is the linked direction finding interaction vector. For example, the numerical feature vector may be represented as [30° quantization value, 10 nautical mile quantization value, 45° quantization value, 12 nautical mile quantization value, 60° quantization value, 15 nautical mile quantization value] (the quantization value in the embodiment of the present application is a numerical value after specific processing, such as converting the angle and distance into a numerical value that is easy to calculate according to the normalization rule).

[0014] For another example, the dynamic direction finding quantization result is the result obtained after processing the direction finding information of the first notification device. It is part of the linked direction finding interaction vector. This result converts the original direction finding information into a quantitative representation that can reflect the dynamic characteristics of the direction finding information, facilitating subsequent calculations and processing. For the direction finding information of platform A, the azimuth is 30° and the distance is 10 nautical miles. For example, the azimuth is quantized according to intervals of 15°, 30° is in the second interval, and is quantized to 2; the distance is quantized according to intervals of 5 nautical miles, 10 nautical miles is in the second interval, and is quantized to 2. Then the dynamic direction finding quantization result of platform A's direction finding information can be [2, 2]. If there is direction finding information from multiple platforms, there will be multiple sub-results mentioned above that together constitute the dynamic direction finding quantization result part.

[0015] In a multi-platform joint search and rescue scenario, to achieve accurate directional positioning, first, in step 110, feature mining is performed on the source direction finding interaction data to obtain a linked direction finding interaction vector. Source direction finding interaction data is crucial in this scenario, describing the joint search and rescue direction finding records for the area where the target notification device is located. For example, in a designated sea area where several platforms are participating in a search and rescue operation, the direction finding data recorded by these platforms for the target notification device (such as an emergency beacon on a wrecked vessel) constitutes the source direction finding interaction data. This data includes the direction finding information of the first notification device, which represents the target notification device.

[0016] The direction-finding information of the first notification device in the embodiments of this application may include data in multiple aspects. For example, the azimuth angle obtained by a certain platform for direction-finding of the target notification device is 30 degrees, and the distance is estimated to be 10 nautical miles, etc. Through feature mining technology, these original direction-finding information of the first notification device is transformed into a dynamic direction-finding quantization result that is more convenient for subsequent processing. This dynamic direction-finding quantization result can be understood as reorganizing and encoding the scattered direction-finding information to make it a vector element that can reflect the dynamic characteristics of the direction-finding information. These elements together constitute a linked direction-finding interaction vector.

[0017] Step 120: Obtain the target multi-platform collaborative direction-finding code; the target multi-platform collaborative direction-finding code is obtained by integrating the state monitoring description vector of the first source state monitoring data on the basis of the dynamic direction-finding quantization result of the direction-finding information of the first notification device.

[0018] Among them, the first source state monitoring data represents the real-time monitoring state of the target notification device.

[0019] For example, the target multi-platform collaborative direction-finding code is a code obtained by integrating the state monitoring description vector of the first source state monitoring data on the basis of the dynamic direction-finding quantization result of the direction-finding information of the first notification device. It comprehensively combines the direction-finding information and the real-time monitoring state information of the target notification device. For example, the first source state monitoring data of the target notification device includes signal strength (three levels: strong, medium, weak, for example, the current is medium, quantified as 1) and battery power (100%, for example, the current battery power is 80%, quantified as 4 according to each 20% as an interval). The previously obtained dynamic direction-finding quantization results are, for example, [2, 2] for platform A and [3, 3] for platform B, etc. Then the target multi-platform collaborative direction-finding code can integrate this information, for example, as [2, 2, 1, 4, 3, 3, 1, 4] (in the embodiments of this application, the first two elements are the dynamic direction-finding quantization results of platform A, the middle two elements are the quantified values of the source state monitoring data corresponding to platform A, and the last four elements are the corresponding information of platform B).

[0020] Another example is that the state monitoring description vector is a vector used to describe the first source state monitoring data. The first source state monitoring data represents the real-time monitoring state of the target notification device, and the state monitoring description vector represents these state information in a set quantization and arrangement manner. As mentioned above, the signal strength (quantified as 1) and battery power (quantified as 4) of the target notification device, the state monitoring description vector may be [1, 4]. If there is other state monitoring information, such as device temperature (for example, within the normal temperature range, quantified as 0), then the state monitoring description vector can be [1, 4, 0].

[0021] In step 120, the obtained target multi-platform collaborative direction-finding code is obtained by integrating the state monitoring description vector of the first source state monitoring data on the basis of the dynamic direction-finding quantization result of the direction-finding information of the first notification device. The first source state monitoring data characterizes the real-time monitoring state of the target notification device. For example, data such as the power state of the target notification device (such as 80% remaining power) and the fluctuation of the signal strength (such as the change trend of the signal strength from strong to weak over a period of time) belong to the first source state monitoring data. These data are represented in the form of a state monitoring description vector and then integrated with the previously obtained dynamic direction-finding quantization result of the direction-finding information of the first notification device, so as to obtain the target multi-platform collaborative direction-finding code. This integration method can be understood as bundling the "position-related information (direction-finding information)" and "state information" of the target notification device together to form a more comprehensive code, so that more factors can be comprehensively considered in subsequent processing.

[0022] Step 130: Determine the target linkage direction-finding interaction vector through the dynamic direction-finding quantization result of the direction-finding information of the first notification device in the linkage direction-finding interaction vector and the target multi-platform collaborative direction-finding code.

[0023] For example, the target linkage direction-finding interaction vector is a vector determined by the dynamic direction-finding quantization result of the direction-finding information of the first notification device in the linkage direction-finding interaction vector and the target multi-platform collaborative direction-finding code. It further integrates the direction-finding information and the state information, and is a vector that more comprehensively reflects the situation of the target notification device. For example, the dynamic direction-finding quantization result in the linkage direction-finding interaction vector is [2, 2, 3, 3] (the quantization results of the direction-finding information of two platforms), and the target multi-platform collaborative direction-finding code is [2, 2, 1, 4, 3, 3, 1, 4]. The target linkage direction-finding interaction vector obtained through a set algorithm (such as weighted summation or other logical relationships) can be [2 + 2×2, 2 + 2×2, 3 + 3×1, 3 + 3×4], that is, [6, 6, 6, 15].

[0024] In step 130, it involves determining the target joint direction-finding interaction vector by means of the dynamic direction-finding quantization result of the direction-finding information of the first reporting device in the joint direction-finding interaction vector and the target multi-platform collaborative direction-finding coding. The embodiments of the present application are based on the results of the previous two steps for further integration and processing. For example, the azimuth information, distance information in the dynamic direction-finding quantization result, and the source state information in the target multi-platform collaborative direction-finding coding, etc., through a set processing method (which can be based on an AI algorithm model, such as the processing logic of the multi-layer perceptron part in a neural network model), these information are fused and adjusted, and finally the target joint direction-finding interaction vector is determined. This target joint direction-finding interaction vector contains richer and more accurate information, covering both the dynamic quantization result related to the direction-finding of the target reporting device and its real-time status monitoring information, making the description of the target reporting device more comprehensive.

[0025] In the complex scenario of multi-platform joint search and rescue, step 130 aims to determine the target joint direction-finding interaction vector by means of the dynamic direction-finding quantization result of the direction-finding information of the first reporting device in the joint direction-finding interaction vector and the target multi-platform collaborative direction-finding coding. This process involves the deep fusion and processing of multi-source information to generate a vector that more comprehensively and accurately reflects the state and position relationship of the target reporting device.

[0026] It can be understood that the dynamic direction-finding quantization result in the joint direction-finding interaction vector is obtained after feature mining of the direction-finding information of the first reporting device. Taking maritime search and rescue as an example, for instance, there are three platforms participating in the direction-finding of a target reporting device (such as a distress beacon on a shipwreck). The initial direction-finding information of platform A may include an azimuth of 30° and a distance of 10 nautical miles from the target reporting device; the azimuth of platform B is 45° and the distance is 12 nautical miles; the azimuth of platform C is 60° and the distance is 15 nautical miles. When performing dynamic direction-finding quantization, for the azimuth, a quantization rule can be set. For example, taking 15° as a quantization interval, then the azimuth of 30° of platform A is in the second interval and is quantized to 2; for the distance, taking 5 nautical miles as a quantization interval, 10 nautical miles is in the second interval and is quantized to 2. Similarly, platforms B and C are also quantized according to the above rules. The finally obtained dynamic direction-finding quantization result can be a numerical feature vector, such as [2, 2, 3, 3, 4, 4] (the embodiments of the present application respectively correspond to the azimuth and distance quantization results of platforms A, B, and C). This dynamic direction-finding quantization result represents the direction-finding information of each platform for the target reporting device in a concise and computationally convenient manner, reflecting the quantization characteristics of the azimuth and distance of the target reporting device relative to each platform.

[0027] The target multi-platform collaborative direction finding coding is obtained by integrating the state monitoring description vector of the first source state monitoring data on the basis of the dynamic direction finding quantization result. The first source state monitoring data characterizes the real-time monitoring state of the target reporting device. For example, the signal strength of the target reporting device can be divided into three levels: strong, medium, and weak. For instance, if the current signal strength is medium, it is quantized to 1 according to a certain quantization rule; for the battery level, if it is initially 100% and the current battery level is 80%, with each 20% as a quantization interval, it is quantized to 4. If there is other state monitoring information, such as the device temperature being normal (quantized to 0), then the state monitoring description vector can be [1, 4, 0]. Integrating this state monitoring description vector with the previous dynamic direction finding quantization result, for example, the dynamic direction finding quantization results of platforms A, B, and C are [2, 2, 3, 3, 4, 4], then the target multi-platform collaborative direction finding coding can be [2, 2, 1, 4, 0, 3, 3, 1, 4, 0, 4, 4, 1, 4, 0] (in the embodiments of this application, the dynamic direction finding quantization results of each platform and the corresponding state monitoring description vectors are arranged in sequence). This coding integrates the direction finding information and the state information of the target reporting device, providing a rich information basis for subsequent determination of the target linkage direction finding interaction vector.

[0028] In the embodiments of this application, an exemplary implementation manner of the process of determining the target linkage direction finding interaction vector can be implemented based on a weighting algorithm. For example, for the azimuth quantization value part in the dynamic direction finding quantization result, since the azimuth is of great significance in determining the target position, a higher weight can be given. For example, the weight is 0.6, while for the distance quantization value part, the weight is set to 0.3, for the signal strength part in the state monitoring description vector, the weight is set to 0.05, and for the battery level part, the weight is set to 0.05. Taking platform A as an example, its azimuth quantization value is 2, the distance quantization value is 2, the signal strength quantization value is 1, and the battery level quantization value is 4. The result calculated according to the weighting algorithm is: for the azimuth part, it is 2×0.6 = 1.2, for the distance part, it is 2×0.3 = 0.6, for the signal strength part, it is 1×0.05 = 0.05, and for the battery level part, it is 4×0.05 = 0.2. Combining these results in a certain order, a new sub-vector is obtained for platform A, for example, [1.2, 0.6, 0.05, 0.2]. Using the same method to calculate for platforms B and C, and then combining these sub-vectors to obtain the target linkage direction finding interaction vector.

[0029] In addition to the weighted algorithm, other algorithm models can also be used to determine the target linkage direction-finding interaction vector. For example, a method based on cluster analysis can be adopted. The dynamic direction-finding quantization results and the target multi-platform collaborative direction-finding coding are regarded as a data set with multiple features. The data points are clustered according to the similarity of the direction-finding information and the status information. For example, the platform data points with similar azimuth angles, similar distances, and similar status information (such as similar signal strength and power) are clustered into one category. Then, key information is extracted from each cluster to construct the target linkage direction-finding interaction vector. For example, the direction-finding information and status information of the platform closest to the target reporting device in a cluster are selected as the representative information of the cluster, and the representative information of each cluster is combined to form the target linkage direction-finding interaction vector.

[0030] In addition, a processing method based on a neural network model can also be considered. The dynamic direction-finding quantization results and the target multi-platform collaborative direction-finding coding can be used as the input layer data of the neural network. The hidden layer of the neural network can mine the complex relationships between the input data through non-linear transformation of the input data. For example, the neurons in the hidden layer can adaptively adjust the weights according to the correlations between the direction-finding information and status information of different platforms. Finally, the target linkage direction-finding interaction vector is output through the output layer. This method can automatically learn the internal laws in the data, so as to more accurately determine the target linkage direction-finding interaction vector.

[0031] In an actual maritime search and rescue scenario, multiple search and rescue platforms are looking for a wrecked ship. There are certain errors and uncertainties in the direction-finding information of each platform and the status information of the target reporting device. By fusing the above-mentioned multi-source information to determine the target linkage direction-finding interaction vector, these errors and uncertainties can be effectively reduced. For example, the direction-finding information of platform A may be biased due to the influence of the marine environment (such as sea waves, magnetic field interference, etc.), and the status monitoring data of platform B may not be accurate enough due to the accuracy problem of the device itself. However, by fusing the information of multiple platforms and using the comprehensive information of azimuth angle and distance in the dynamic direction-finding quantization results, as well as the status information in the target multi-platform collaborative direction-finding coding, they can complement and correct each other. The finally obtained target linkage direction-finding interaction vector can more accurately reflect the true position and status of the target reporting device.

[0032] In the embodiments of the present application, the target linkage direction-finding interaction vector can contain more comprehensive, accurate, and deeply integrated information about the target reporting device. This vector not only reflects the azimuth and distance relationship of the target reporting device relative to each platform but also includes the real-time status information of the target reporting device. For example, it can clearly represent the approximate azimuth angle, distance, signal strength, power, and other status information of the target reporting device at a certain moment. In this way, the entire technical solution can effectively improve the accuracy and reliability of the positioning of the target reporting device in the multi-platform joint search and rescue scenario, greatly improving the efficiency of the maritime formation search and rescue.

[0033] In addition, the method for determining the target linkage direction-finding interaction vector fully demonstrates the advantages of modern technology in processing multi-source and complex information. By integrating data from different sources and of different types, the scattered and potentially error-prone information is transformed into a vector that can accurately describe the target. The algorithm selection and parameter settings in this process can be adjusted and optimized according to the actual application scenario and data characteristics to adapt to different search and rescue environments and the characteristics of the target reporting device, thereby continuously improving the performance of the entire positioning technology.

[0034] In actual operation, to further improve the accuracy of the target linkage direction-finding interaction vector, data preprocessing and postprocessing can also be performed. In the data preprocessing stage, the direction-finding information and status monitoring data can be cleaned to remove outliers. For example, if a platform suddenly measures an azimuth angle or distance value that is significantly different from other platforms, which may be caused by equipment failure or temporary interference, the outlier can be corrected or removed. In the postprocessing stage, the determined target linkage direction-finding interaction vector can be verified and adjusted. For example, cross-verify the target linkage direction-finding interaction vector using historical data or other auxiliary information. If a large deviation is found from existing reliable data or empirical rules, the parameters or algorithms in the calculation process can be adjusted to re-determine the target linkage direction-finding interaction vector.

[0035] It can be seen that step 130 determines the target linkage direction-finding interaction vector by fusing the dynamic direction-finding quantization result and the target multi-platform collaborative direction-finding coding in the linkage direction-finding interaction vector. This process involves various algorithms and data processing techniques and is of irreplaceable importance in the multi-platform joint search and rescue scenario, providing a key information integration and optimization function for the entire positioning technology solution.

[0036] Step 140: Generate a spatio-temporal orientation positioning label corresponding to the regional joint search and rescue direction-finding record according to the target linkage direction-finding interaction vector.

[0037] For example, a spatiotemporal directional positioning tag is generated based on the target linkage direction finding interaction vector. It identifies the spatiotemporal location and status of the target notification device corresponding to the regional joint search and rescue direction finding record. For example, the target linkage direction finding interaction vector indicates the location of the target notification device at a specific moment (e.g., 10:00 on July 1, 2023). The calculated azimuth is 40°, the distance is 13 nautical miles, the status is medium signal strength, and the battery level is 80%. Therefore, the spatiotemporal directional positioning tag might be "2023-07-01 10:00, azimuth 40°, distance 13 nautical miles, medium signal strength, battery level 80%."

[0038] In step 140, a spatiotemporal directional positioning tag corresponding to the regional joint search and rescue direction-finding record is generated based on the target-linked direction-finding interaction vector. This spatiotemporal directional positioning tag can be understood as a precise "identity tag" for the target notification device, accurately identifying its location and status in both time and space. For example, the tag can clearly indicate that the target notification device was located at a specific longitude and latitude (e.g., 120 degrees east longitude, 30 degrees north latitude) at a specific moment (e.g., 10:30 a.m. on July 1, 2023). The tag also includes its signal source status (e.g., 80% battery remaining, medium signal strength, etc.). This spatiotemporal directional positioning tag enables participating platforms in multi-platform joint search and rescue scenarios to more accurately locate the target notification device, whether it is a ship at sea or other distressed target, thereby significantly improving the efficiency of maritime formation search and rescue.

[0039] As can be understood, the above technical solution achieves a gradual progression and continuous information integration. Starting from the initial source direction-finding interaction data, the linked direction-finding interaction vector is derived by mining its features. Then, the source status monitoring data is integrated to obtain the target multi-platform collaborative direction-finding code. The target linked direction-finding interaction vector is then determined, ultimately generating a spatiotemporal directional positioning tag. Each step is closely linked, fully utilizing the direction-finding information from multiple platforms and the status information of the target notification equipment, effectively improving the accuracy and reliability of locating distressed objects during maritime search and rescue.

[0040] In an actual maritime search and rescue scenario, for example, there are three search and rescue platforms, A, B, and C. Platform A's initial direction-finding information for the target notification device is an azimuth of 45 degrees and a distance of 12 nautical miles; Platform B's initial direction-finding information is an azimuth of 60 degrees and a distance of 15 nautical miles; and Platform C's initial direction-finding information is an azimuth of 50 degrees and a distance of 13 nautical miles. This initial direction-finding information is part of the first notification device's direction-finding information in the source direction-finding interaction data. Through feature mining, the resulting dynamic direction-finding quantization results may convert this information into a format more suitable for computer processing, such as expressing the relative position of each platform relative to the target notification device in the form of a coordinate vector.

[0041] Then, for the status monitoring data of the target notification device, such as detecting that its signal emission frequency has slight fluctuations over a period of time and the power has dropped from the initial 100% to 90%, these status monitoring data are integrated with the previous dynamic direction finding quantization results in the form of a status monitoring description vector to obtain the target multi-platform collaborative direction finding code.

[0042] When determining the target linkage direction finding interaction vector, comprehensively consider the position relationship information in the dynamic direction finding quantization results of each platform and the status information in the target multi-platform collaborative direction finding code. For example, according to the relative position relationship of platforms A, B, and C and the status information of the target notification device, the information of each platform is fused in a way similar to weighted averaging to obtain the target linkage direction finding interaction vector.

[0043] Finally, the spatio-temporal orientation positioning label generated based on this target linkage direction finding interaction vector can accurately indicate the position and status of the target notification device at a specific moment. For example, at 11:00 on July 1, 2023, the target notification device is located at 121 degrees east longitude and 31 degrees north latitude, with a fluctuating signal emission frequency and 90% remaining power. Such accurate positioning and status identification are very crucial for maritime search and rescue, enabling the search and rescue team to carry out rescue work more efficiently, reducing the waiting time of the distressed target for rescue, and improving the success rate of rescue.

[0044] It can be seen that the above technical solution with the dynamic limit orientation positioning system as the execution subject, through the gradual processing and information integration of the source direction finding interaction data, has successfully achieved the precise orientation and positioning of the target notification device in the multi-platform joint search and rescue scenario, providing an efficient and reliable solution for similar scenarios such as maritime search and rescue.

[0045] In an optional embodiment, the obtaining of the target multi-platform collaborative direction finding code in step 120 includes: determining the platform collaborative direction finding knowledge graph in the platform database; the platform collaborative direction finding knowledge graph includes platform collaborative direction finding knowledge corresponding to multiple second notification device direction finding information respectively; screening the platform collaborative direction finding knowledge corresponding to the first notification device direction finding information from the platform collaborative direction finding knowledge graph to obtain the target multi-platform collaborative direction finding code.

[0046] When obtaining the target multi-platform collaborative direction-finding code, determining the platform collaborative direction-finding knowledge graph in the platform database is a key starting point. The platform collaborative direction-finding knowledge graph contains platform collaborative direction-finding knowledge corresponding to the direction-finding information of multiple second reporting devices respectively. The platform collaborative direction-finding knowledge graph in the embodiments of the present application is a comprehensive knowledge system, which integrates and correlates the direction-finding related knowledge of different reporting devices. For example, there are multiple reporting devices such as reporting device 1, reporting device 2, and reporting device 3 (in the embodiments of the present application, different reporting devices are used to distinguish the second reporting devices). For the direction-finding information of the second reporting device of reporting device 1, it may include information such as the azimuth angle is 40° and the distance is 12 nautical miles. The corresponding platform collaborative direction-finding knowledge graph contains the relevant platform collaborative direction-finding knowledge, which is obtained through a series of processing processes, including the integration of multi-faceted information such as the status monitoring data of the reporting device.

[0047] Screening the platform collaborative direction-finding knowledge corresponding to the direction-finding information of the first reporting device from the platform collaborative direction-finding knowledge graph to obtain the target multi-platform collaborative direction-finding code is similar to finding and extracting the knowledge fragments related to a specific target in a huge knowledge warehouse. For example, in the direction-finding information of the first reporting device, the azimuth angle is 30° and the distance is 10 nautical miles. According to this direction-finding information, find the platform collaborative direction-finding knowledge that matches or has the highest correlation degree in the platform collaborative direction-finding knowledge graph. This knowledge may include the processing results of previous similar direction-finding information, relevant status monitoring data and other multi-faceted information. These information are integrated to form the target multi-platform collaborative direction-finding code.

[0048] Further, the steps of generating the platform collaborative direction-finding knowledge corresponding to the direction-finding information of the second reporting device include: obtaining multiple second source status monitoring data of the reporting device characterized by the direction-finding information of the second reporting device; respectively performing spatio-temporal element identification on the multiple second source status monitoring data to obtain multiple spatio-temporal element status vectors, and performing perturbation feature detection on the direction-finding information of the second reporting device to obtain the perturbation state detection vector of the reporting device; performing feature integration based on the perturbation state detection vector and the multiple spatio-temporal element status vectors to obtain the platform collaborative direction-finding knowledge corresponding to the direction-finding information of the second reporting device.

[0049] In the process of generating the platform collaborative direction-finding knowledge corresponding to the direction-finding information of the second reporting device, first obtain multiple second source status monitoring data of the reporting device characterized by the direction-finding information of the second reporting device. Taking reporting device 1 as an example, its second source status monitoring data may include data such as signal strength, power, and device temperature. For example, the signal strength is medium (which can be quantified as 1), the power is 80% (quantified as 4 according to each 20% as an interval), and the device temperature is normal (quantified as 0).

[0050] Then, spatiotemporal element identification is respectively performed on multiple pieces of second source state monitoring data to obtain multiple spatiotemporal element state vectors. For example, for the second source state monitoring data of the power consumption, the spatiotemporal element identification may consider the change trend of the power consumption over time and the influence of different spatial positions (such as the influence of the distances of different platforms from the notification device on the power consumption monitoring). For example, the spatiotemporal element state vector obtained after spatiotemporal element identification is [1, 2], where 1 may represent a certain feature of the power consumption in the time dimension, and 2 represents a certain feature in the spatial dimension. Similarly, spatiotemporal element identification is also performed on the signal strength and the device temperature to obtain their respective spatiotemporal element state vectors.

[0051] Next, perturbation feature detection is performed on the direction finding information of the second notification device to obtain the perturbation state detection vector of the notification device. For example, the azimuth angle of the direction finding information of notification device 1 is 40°, and the distance is 12 nautical miles. During the measurement process, it may be perturbed by the marine environment (such as sea waves, magnetic fields, etc.). The perturbation state detection vector obtained after perturbation feature detection can be [3, 1] (in Embodiment 3 of this application, 3 may represent the degree of perturbation of the azimuth angle, and 1 represents the degree of perturbation of the distance).

[0052] In the next step, the feature integration based on the perturbation state detection vector and the multiple spatiotemporal element state vectors to obtain the platform collaborative direction finding knowledge corresponding to the direction finding information of the second notification device includes: for each spatiotemporal element state vector in the multiple spatiotemporal element state vectors, integrating the spatiotemporal element state vector on the basis of the perturbation state detection vector to obtain a collaborative direction finding cross vector; processing each of the collaborative direction finding cross vectors to obtain the platform collaborative direction finding knowledge of the notification device.

[0053] In this embodiment, the platform collaborative direction-finding knowledge corresponding to the direction-finding information of the second notification device can be obtained through feature integration based on the disturbance state detection vector and multiple spatio-temporal element state vectors. Specifically, for each spatio-temporal element state vector among the multiple spatio-temporal element state vectors, the spatio-temporal element state vector is integrated on the basis of the disturbance state detection vector to obtain a collaborative direction-finding cross vector. Taking the spatio-temporal element state vector [1, 2] corresponding to the previous power and the disturbance state detection vector [3, 1] as an example, the collaborative direction-finding cross vector obtained after integration can be [3 + 1, 1 + 2] = [4, 3]. The platform collaborative direction-finding knowledge of the notification device is obtained by processing each collaborative direction-finding cross vector. This processing process may involve operations such as weighted summation and feature screening of multiple collaborative direction-finding cross vectors. For example, if there are three collaborative direction-finding cross vectors [4, 3], [2, 5], and [3, 4] respectively, the platform collaborative direction-finding knowledge obtained through weighted summation (such as weights 0.3, 0.3, and 0.4 respectively) can be [(4×0.3 + 2×0.3 + 3×0.4), (3×0.3 + 5×0.3 + 4×0.4)] = [3, 4].

[0054] In this way, by constructing a platform collaborative direction-finding knowledge graph and screening the target multi-platform collaborative direction-finding code therefrom, the existing platform collaborative direction-finding knowledge can be fully utilized to improve the accuracy and efficiency of code acquisition. In the process of generating the platform collaborative direction-finding knowledge, the spatio-temporal elements of the source state monitoring data of the notification device and the disturbance characteristics of the direction-finding information are comprehensively considered, so that the platform collaborative direction-finding knowledge can more comprehensively and accurately reflect the actual situation of the notification device. This way of feature integration can effectively integrate information from multiple aspects, avoid one-sidedness of information, thereby improving the positioning accuracy and reliability of the target notification device in the entire multi-platform joint search and rescue process, and is of great significance for improving the efficiency of maritime search and rescue, etc.

[0055] In a preferred embodiment, the obtaining of the target multi-platform collaborative direction-finding code includes: performing spatio-temporal element identification on the second source state monitoring data of the target notification device to obtain the spatio-temporal element state vector of the target notification device; performing disturbance feature detection on the direction-finding information of the first notification device to obtain the disturbance state detection vector of the target notification device; and integrating the spatio-temporal element state vector of the target notification device on the basis of the disturbance state detection vector of the target notification device to obtain the target multi-platform collaborative direction-finding code.

[0056] In the embodiment of the present application, first, spatio-temporal element recognition is performed on the second source state monitoring data of the target notification device to obtain the spatio-temporal element state vector of the target notification device. The second source state monitoring data contains various aspects of state information related to the target notification device. For example, for the target notification device, regarding the signal strength information in its second source state monitoring data, if the signal strength is divided into different levels, such as strong corresponding to the value 5, medium corresponding to the value 3, and weak corresponding to the value 1. For instance, if the signal strength of the target notification device is medium, then the corresponding value of the signal strength element in the spatio-temporal element state vector is 3. Looking at the battery level, if the full battery level is 100% and it is quantified in intervals of 20%, when the battery level is 60%, it is in the 3rd interval and the corresponding value is 3. Regarding the device temperature, for example, the normal temperature range is a numerical interval, and the corresponding value within this interval is 0. Combining this information, the spatio-temporal element state vector of the target notification device can be [3, 3, 0] (in other cases, it may contain more elements).

[0057] Next, perturbation feature detection is performed on the direction-finding information of the first notification device to obtain the perturbation state detection vector of the target notification device. The direction-finding information of the first notification device contains direction-finding related information such as the azimuth angle and distance of the target notification device. For example, for the measurement of the azimuth angle, it may be affected by various factors. For instance, the measured value of the azimuth angle is 45°. Due to the influence of the surrounding environment (such as magnetic field interference, etc.), a certain perturbation of this azimuth angle is detected. If the perturbation degree is divided into different levels, slight perturbation corresponds to the value 1, medium perturbation corresponds to the value 2, and severe perturbation corresponds to the value 3. For example, if the perturbation of this azimuth angle is slight perturbation, then the corresponding value of the azimuth angle in the perturbation state detection vector is 1. For the distance measurement, for example, the measured distance value is 10 nautical miles. If there is a certain measurement error, the perturbation levels are divided according to the error magnitude. If it is a slight error, the corresponding value is 1. Then the obtained perturbation state detection vector of the target notification device can be [1, 1].

[0058] Finally, based on the perturbation state detection vector of the target notification device, the spatio-temporal element state vector of the target notification device is integrated to obtain the target multi-platform collaborative direction-finding code. Taking the previously obtained spatio-temporal element state vector [3, 3, 0] and perturbation state detection vector [1, 1] as an example, the integration method can add the corresponding elements or combine them according to a certain rule. If it is the simple addition method, the obtained target multi-platform collaborative direction-finding code is [1 + 3, 1 + 3, 0] = [4, 4, 0]. This integration method is not a simple patchwork, but comprehensively considers the perturbation situation of the direction-finding information of the target notification device and the spatio-temporal elements of the state monitoring data, making the target multi-platform collaborative direction-finding code contain richer and more comprehensive information.

[0059] Thus, by identifying the spatio-temporal elements of the second source status monitoring data of the target notification device and detecting the perturbation characteristics of the direction-finding information of the first notification device, and then integrating them to obtain the target multi-platform collaborative direction-finding code, this method can comprehensively consider various situations of the target notification device. It includes both the spatio-temporal elements of the status monitoring data and the perturbation characteristics of the direction-finding information, enabling the target multi-platform collaborative direction-finding code to more accurately reflect the actual situation of the target notification device. This helps to improve the accuracy and reliability of locating the target notification device in scenarios such as multi-platform joint search and rescue, and thus enhances the efficiency of the entire search and rescue and related work.

[0060] In the specific implementation process, generating the spatio-temporal orientation positioning label corresponding to the regional joint search and rescue direction-finding record according to the target linkage direction-finding interaction vector described in step 140 includes: generating an orientation positioning vector with adjustable limits according to the first orientation positioning drift label; inputting the target linkage direction-finding interaction vector and the orientation positioning vector with adjustable limits into a target generative adversarial model for spatio-temporal phase attention enhancement to obtain a spatio-temporal phase attention vector; identifying the spatio-temporal phase attention vector to generate the spatio-temporal orientation positioning label corresponding to the regional joint search and rescue direction-finding record; wherein, the target generative adversarial model is obtained by improving a pre-tuned generative adversarial model based on the collaborative direction-finding cross vector of historical notification devices, and the collaborative direction-finding cross vector of historical notification devices is the result of feature integration of the second source status monitoring data of the historical notification devices and the direction-finding information of the second notification devices of the historical notification devices.

[0061] In the embodiment of the present application, first, an orientation positioning vector with adjustable limits is generated according to the first orientation positioning drift label. The first orientation positioning drift label contains certain initial information related to positioning, and an orientation positioning vector with adjustable limits is constructed starting from this label. For example, the first orientation positioning drift label may contain information such as the initial azimuth range of the target notification device. If the range is 30° - 40°, this information will be converted into elements in the vector when generating the orientation positioning vector with adjustable limits. For example, the orientation positioning vector with adjustable limits is expressed as [30, 40, other elements] (the other elements in the embodiment of the present application can be the quantitative representation of other information related to positioning, such as the distance range, etc.).

[0062] Next, the target linkage direction-finding interaction vector and the limit-adjustable orientation positioning vector are input into the target generative adversarial model for spatio-temporal phase attention enhancement to obtain the spatio-temporal phase attention vector. The target linkage direction-finding interaction vector is a vector containing multi-faceted information integrated through a series of previous steps. For example, it may contain the processed results of direction-finding information, status monitoring information, etc. of the target reporting device. For example, the target linkage direction-finding interaction vector is [10, 12, 3, 4] (the values in the embodiments of this application respectively represent the quantization values of different direction-finding and status monitoring information). This vector and the previous limit-adjustable orientation positioning vector [30, 40, other elements] are input into the target generative adversarial model together.

[0063] Among them, the target generative adversarial model is obtained by improving the pre-debugged generative adversarial model based on the collaborative direction-finding cross vector of historical reporting devices. The collaborative direction-finding cross vector of historical reporting devices is the result of feature integration of the second source status monitoring data of historical reporting devices and the second reporting device direction-finding information of historical reporting devices. For example, for a certain historical reporting device, the signal strength in its second source status monitoring data is medium (quantified as 3), the battery power is 80% (quantified as 4), and the device temperature is normal (quantified as 0); the azimuth angle in its second reporting device direction-finding information is 45° (quantified as 5), and the distance is 12 nautical miles (quantified as 3). The collaborative direction-finding cross vector obtained through feature integration can be [3 + 5, 4 + 3, 0] = [8, 7, 0]. The pre-debugged generative adversarial model is improved based on the collaborative direction-finding cross vectors of multiple above-mentioned historical reporting devices, so that the target generative adversarial model can better adapt to the actual situation.

[0064] In the target generative adversarial model, the input target linkage direction-finding interaction vector and the limit-adjustable orientation positioning vector are processed through the mechanism of spatio-temporal phase attention enhancement to obtain the spatio-temporal phase attention vector. This vector is obtained through complex operations inside the model. It synthesizes the information of the input vectors and enhances the spatio-temporal phase information under the action of the model. For example, the obtained spatio-temporal phase attention vector can be [15, 18, 5, 6] (the values in the embodiments of this application are the results after model processing, related to the input vectors and reflecting the characteristics after spatio-temporal phase attention enhancement).

[0065] Finally, identify the spatio-temporal orientation attention vector to generate spatio-temporal orientation and positioning labels corresponding to the regional joint search and rescue direction finding records. The identification process interprets the elements in the spatio-temporal orientation attention vector according to pre-set rules. For example, if the first element in the spatio-temporal orientation attention vector represents a certain quantization result of the azimuth angle, the second element represents the quantization result of the distance, the third element represents the quantization result related to the signal strength, and the fourth element represents the quantization result related to the power, then the spatio-temporal orientation and positioning label generated based on these elements can be "the angle represented by the corresponding value 15 of the azimuth angle, the distance represented by the corresponding value 18 of the distance, the strength represented by the corresponding value 5 of the signal strength, and the power state represented by the corresponding value 6 of the power".

[0066] In this way, by generating an orientation and positioning vector with adjustable limits based on the first orientation and positioning drift label, basic positioning-related information is provided for subsequent operations. The spatio-temporal orientation attention is strengthened using the target generative adversarial model, and the model is improved with the collaborative direction finding cross vector of the historical notification device, enabling the model to better process the target-linked direction finding interaction vector and the orientation and positioning vector with adjustable limits, improving the accuracy and effectiveness of the processing. The finally generated spatio-temporal orientation and positioning label can accurately reflect the relevant information of the regional joint search and rescue direction finding records, thus improving the accuracy and reliability of the positioning of the target notification device in scenarios such as multi-platform joint search and rescue.

[0067] In the embodiment of the present application, the steps of obtaining the target generative adversarial model include: performing spatio-temporal element identification and perturbation processing on the first historical source state monitoring data to obtain an orientation and positioning vector with adjustable limits; integrating the dynamic direction finding quantization result of the direction finding information of the historical notification device with the spatio-temporal element state vector of the second historical source state monitoring data to obtain a historical collaborative direction finding cross vector; the first historical source state monitoring data and the second historical source state monitoring data are the source state monitoring data of the same historical notification device, and the direction finding information of the historical notification device is the direction finding interaction data representing the historical notification device; inputting the orientation and positioning vector with adjustable limits and the historical collaborative direction finding cross vector into the pre-debugged generative adversarial model for spatio-temporal orientation attention strengthening to obtain a target spatio-temporal element state vector; identifying the target spatio-temporal element state vector to generate source state monitoring training data; and improving the model weights of the pre-debugged generative adversarial model based on the source state monitoring training data to obtain the target generative adversarial model.

[0068] In the embodiment of the present application, first, the spatio-temporal elements of the first historical source state monitoring data are identified and perturbed to obtain a history-oriented positioning vector with adjustable limits. The first historical source state monitoring data contains state-related information of historical notification devices at different spatio-temporal points. For example, the signal strength in the first historical source state monitoring data of a certain historical notification device, if quantified according to a set quantization standard, a strong signal strength is quantified as 5, and when the signal strength is monitored as medium at a certain moment, it is quantified as 3. Considering the influence of different spatial positions on the signal strength (such as the distance from the receiving device), after identifying the spatio-temporal elements, and adding some possible perturbation factors (such as environmental interference), if the perturbation effect is small, it is quantified as 1. Combining this information, a part of the elements in the history-oriented positioning vector with adjustable limits is obtained. Looking at the battery level, for example, a full battery is 100%, quantified in intervals of 20%. When the battery level is 60%, it is in the 3rd interval, corresponding to the value 3. If there is a certain fluctuation perturbation in the battery level monitoring, it is quantified as 1, which also becomes an element of the history-oriented positioning vector with adjustable limits. For example, the finally obtained history-oriented positioning vector with adjustable limits is [3 + 1, 3, other elements] (the other elements in the embodiment of the present application can be the results of similar processing of other state monitoring information).

[0069] Next, the dynamic direction-finding quantization result of the direction-finding information of the historical notification device is integrated with the spatio-temporal element state vector of the second historical source state monitoring data to obtain a historical collaborative direction-finding cross vector. The direction-finding information of the historical notification device represents the direction-finding interaction data of the historical notification device. For example, the azimuth angle in the direction-finding information of the historical notification device is 45°, quantified as 5 according to a certain quantization rule, and the distance is 12 nautical miles, quantified as 3. For the second historical source state monitoring data, such as the signal strength is medium (quantified as 3), the battery level is 80% (quantified as 4), and the device temperature is normal (quantified as 0). After identifying the spatio-temporal elements, the spatio-temporal element state vector is [3, 4, 0]. Integrating the dynamic direction-finding quantization result [5, 3] of the direction-finding information of the historical notification device with the spatio-temporal element state vector [3, 4, 0], it can be obtained by adding elements or other specific rules to get the historical collaborative direction-finding cross vector, for example, [5 + 3, 3 + 4, 0] = [8, 7, 0].

[0070] Then, input the adjustable-limit historical orientation positioning vector and the historical cooperative direction-finding cross vector into the pre-debugged generative adversarial model to obtain the target spatio-temporal element state vector through spatio-temporal phase attention enhancement. For example, the adjustable-limit historical orientation positioning vector is [4, 3, other elements], and the historical cooperative direction-finding cross vector is [8, 7, 0]. Input these two vectors into the pre-debugged generative adversarial model. Inside the model, these input vectors are processed through the spatio-temporal phase attention enhancement mechanism. This enhancement mechanism focuses on and optimally processes the spatio-temporal phase-related information in the input vectors based on the internal structure and algorithm logic of the model. After being processed by the model, the target spatio-temporal element state vector is obtained. For example, the obtained target spatio-temporal element state vector is [10, 12, 3] (the numerical values in the embodiments of this application are the results of complex model operations, related to the input vectors and reflecting the characteristics after spatio-temporal phase attention enhancement).

[0071] After that, identify the target spatio-temporal element state vector to generate the training data for source state monitoring. The identification process interprets and converts the elements in the target spatio-temporal element state vector according to pre-set rules. For example, if the first element in the target spatio-temporal element state vector represents a certain quantization result related to the azimuth angle, the second element represents a quantization result related to the distance, and the third element represents a quantization result related to the signal strength, then according to these elements, the training data for source state monitoring is generated according to a certain mapping relationship. For example, if the quantization result related to the azimuth angle is 10, it may be converted into a set angle range according to the mapping relationship; the quantization result related to the distance is 12, which is converted into the actual distance range; the quantization result related to the signal strength is 3, which is converted into the corresponding signal strength description. These converted data together constitute the training data for source state monitoring.

[0072] Finally, based on the training data for source state monitoring, improve the model weights of the pre-debugged generative adversarial model to obtain the target generative adversarial model. The training data for source state monitoring contains accurate information related to source state monitoring converted from the target spatio-temporal element state vector. By feeding this information back into the pre-debugged generative adversarial model, the model weights are adjusted. For example, if the training data for source state monitoring indicates that there are deviations between the model's prediction results and the actual results in certain cases, then the model weights are adjusted to optimize the model's prediction ability. After multiple adjustments as described above, the model can better adapt to the actual situation, and finally the target generative adversarial model is obtained.

[0073] Therefore, through a series of operations such as identifying spatio-temporal elements and processing perturbations in historical source state monitoring data, integrating direction-finding information with historical source state monitoring data, and inputting relevant vectors into a pre-debugged generative adversarial model for spatio-temporal phase attention enhancement, the information in historical data can be fully utilized. The generated training data for source state monitoring is used to improve the model weights, and the resulting target generative adversarial model can more accurately process information related to the target reporting device, thereby improving the accuracy and reliability of positioning and enhancing adaptability to different situations in scenarios such as multi-platform joint search and rescue.

[0074] In an alternative technical solution, integrating the dynamic direction-finding quantization result of historical reporting device direction-finding information with the spatio-temporal element state vector of the second historical source state monitoring data to obtain the historical collaborative direction-finding cross vector includes: detecting perturbation features in the historical reporting device direction-finding information to obtain a historical perturbation state detection vector; identifying spatio-temporal elements in the second historical source state monitoring data to obtain a historical spatio-temporal element state vector; and integrating the features of the historical perturbation state detection vector and the historical spatio-temporal element state vector to obtain the historical collaborative direction-finding cross vector.

[0075] In the embodiment of the present application, first, perturbation features are detected in the historical reporting device direction-finding information to obtain a historical perturbation state detection vector. The historical reporting device direction-finding information contains important information related to direction-finding such as the azimuth angle and distance of the historical reporting device. For example, for the direction-finding information of a certain historical reporting device, its azimuth angle is 45°. Due to various possible interference factors, such as magnetic field interference in the environment and reflections from surrounding objects, there may be certain perturbations in the measurement of this azimuth angle. For example, according to the quantization standard for the degree of perturbation, a slight perturbation corresponds to the value 1, a medium perturbation corresponds to the value 2, and a severe perturbation corresponds to the value 3. If the perturbation of this azimuth angle is determined to be a slight perturbation, then the value corresponding to the azimuth angle in the historical perturbation state detection vector is 1. For distance measurement, for example, the measured distance value is 12 nautical miles. If there is a certain measurement error, and the perturbation level is divided according to the error size, if it is a slight error, the corresponding value is 1, then the obtained historical perturbation state detection vector can be [1, 1] (the embodiment of the present application only takes the azimuth angle and distance as examples, and actually may include more perturbation quantizations related to direction-finding information).

[0076] Next, perform spatio-temporal element identification on the second historical source state monitoring data to obtain a historical spatio-temporal element state vector. The second historical source state monitoring data covers various state information of historical notification devices. For example, taking signal strength as an example, if the signal strength is divided into different levels, strong corresponds to the value 5, medium corresponds to the value 3, and weak corresponds to the value 1. For instance, if the signal strength of a historical notification device is medium, then the corresponding value of the signal strength element in the historical spatio-temporal element state vector is 3. Looking at the battery level, if a full battery is 100% and it is quantified in intervals of 20%, when the battery level is 60%, it is in the 3rd interval and the corresponding value is 3. Regarding the device temperature, for example, the normal temperature range is a numerical interval, and the corresponding value within this interval is 0. Combining this information, the historical spatio-temporal element state vector can be [3, 3, 0] (it may actually contain more elements).

[0077] Finally, perform feature integration on the historical disturbance state detection vector and the historical spatio-temporal element state vector to obtain a historical collaborative direction finding cross vector. For example, if the historical disturbance state detection vector is [1, 1] and the historical spatio-temporal element state vector is [3, 3, 0], the feature integration method can add the corresponding elements or combine them according to a certain rule. If it is a simple addition method, the obtained historical collaborative direction finding cross vector is [1 + 3, 1 + 3, 0] = [4, 4, 0]. This feature integration is not a simple patchwork, but combines the disturbance situation of the direction finding information of historical notification devices and the spatio-temporal elements of the source state monitoring data, enabling the historical collaborative direction finding cross vector to more comprehensively and accurately reflect the relevant situation of historical notification devices.

[0078] It can be seen that by performing disturbance feature detection on the direction finding information of historical notification devices, spatio-temporal element identification on the second historical source state monitoring data, and then performing feature integration to obtain a historical collaborative direction finding cross vector, this method can fully consider the disturbance factors in the direction finding process of historical notification devices and the spatio-temporal characteristics of the state monitoring data. The historical collaborative direction finding cross vector can more accurately reflect the actual situation of historical notification devices, which helps to improve the accuracy and effectiveness of data when using historical data for model training or other related operations in the future, and further enhances the reliability and accuracy of the entire technical solution in dealing with similar situations.

[0079] In some examples, the first historical source state monitoring data is different from the second historical source state monitoring data; improving the model weights of the pre-debugged generative adversarial model based on the source state monitoring training data to obtain a target generative adversarial model includes: calculating a first training error based on the comparison result between the source state monitoring training data and the second historical source state monitoring data; using the first training error to improve the model weights of the pre-debugged generative adversarial model to obtain a target generative adversarial model.

[0080] In this example, the first historical signal source status monitoring data and the second historical signal source status monitoring data are different, meaning they monitor the signal source status of the historical notification device from different perspectives or time periods. For example, the first historical signal source status monitoring data may focus on monitoring aspects such as signal strength and device temperature during the early stages of device operation, such as signal strength quantified as 3 (medium strength) and device temperature quantified as 0 (normal temperature range) at a certain moment. The second historical signal source status monitoring data, on the other hand, may focus more on battery power and signal stability during the mid-stage of device operation, such as battery power at 60% (quantified as 3 with a 20% interval) and signal stability quantified as 2 (relative stability).

[0081] Next, the first training error is calculated based on the comparison results between the source state monitoring training data and the second historical source state monitoring data. The source state monitoring training data is obtained through the previous series of operations and contains information related to source state monitoring after the target spatiotemporal element state vector is identified and transformed. For example, the signal strength in the source state monitoring training data is converted to a certain value, such as 4, while the signal strength in the second historical source state monitoring data is 3, resulting in a difference between the two. Difference values are calculated by comparing corresponding elements (such as signal strength, power level, and device temperature), and these differences are combined to obtain the first training error. If there are other elements, such as device temperature, which is 0 in the source state monitoring training data and 0 in the second historical source state monitoring data, the difference for this element is 0 when calculating the first training error. However, the difference in signal strength is 1, and other elements such as power level also have their own differences. These differences are combined according to a certain rule (such as weighted averaging) to obtain the first training error.

[0082] Finally, the first training error is used to improve the model weights of the pre-tuned GAM to obtain the target GAM. The pre-tuned GAM has initial model weights, which determine how the model processes input data and its predictions. Once the first training error is calculated, the model weights can be adjusted accordingly. For example, if the first training error indicates that the model consistently over-predicts signal strength, the weights associated with signal strength prediction can be reduced. Similar adjustments are performed for other factors such as battery level and device temperature. Through these adjustments, the model can better adapt to actual conditions, reduce prediction errors, and ultimately obtain the target GAM.

[0083] In this way, by calculating the first training error using the comparison result between the source state monitoring training data and the second historical source state monitoring data, the difference between the model prediction and the actual historical data can be accurately found. Using this training error to improve the model weights of the pre-debugged generative adversarial model can enable the model to more accurately process tasks related to source state monitoring, which helps to improve the effectiveness of the model in the entire technical solution, enhance the adaptability to different source states, and thus improve the accuracy and reliability of the processing related to the target notification device.

[0084] In yet another alternative embodiment, the spatio-temporal element identification and perturbation processing of the first historical source state monitoring data to obtain a limit-adjustable historical orientation and positioning vector includes: combining the second orientation and positioning drift label with the first historical source state monitoring data to obtain source state drift data; performing spatio-temporal element identification on the source state drift data to obtain a spatio-temporal state drift vector; and performing feature mapping on the spatio-temporal state drift vector to obtain a limit-adjustable historical orientation and positioning vector.

[0085] It can be understood that this embodiment focuses on obtaining a limit-adjustable historical orientation and positioning vector from the first historical source state monitoring data. First, the source state drift data is obtained by combining the second orientation and positioning drift label with the first historical source state monitoring data. The first historical source state monitoring data contains state information related to the historical notification device, such as monitoring information on the signal strength, power, device temperature, etc. of the historical notification device. For example, the signal strength in the first historical source state monitoring data is quantified as 3 (indicating medium strength), the power is 60% (quantified as 3 according to each 20% as an interval), and the device temperature is quantified as 0 (indicating within the normal temperature range). The second orientation and positioning drift label contains information related to positioning drift, such as possible quantitative or qualitative descriptions of the drift direction and drift degree that the historical notification device may have during different times or in different environments. When combining the two, data fusion can be performed according to a certain rule. For example, the drift direction in the second orientation and positioning drift label is associated with the signal strength, and the drift degree is associated with the power, etc. By this combination, the source state drift data is obtained, which is a comprehensive data integrating positioning drift information and the original source state monitoring information. For example, the signal strength in the source state drift data may become [3, drift direction related value], and the power may become [3, drift degree related value], etc.

[0086] Next, perform spatio-temporal element identification on the source state drift data to obtain a spatio-temporal state drift vector. In the source state drift data, since it incorporates information related to positioning drift and source state information, multiple factors need to be comprehensively considered during spatio-temporal element identification. Regarding signal strength, not only the changes of its own over time and space in the original state need to be considered, but also the impact of positioning drift on its spatio-temporal changes. For example, within a certain time period, due to the positioning drift direction towards the signal interference source, the spatial change of the signal strength shows anomalies. According to the rules of spatio-temporal element identification, this spatio-temporal change of the signal strength is quantified. For instance, the representation of the signal strength in the spatio-temporal state drift vector after quantization is [4] (the 4 in the embodiment of this application is the quantization result after considering multiple factors). Regarding the battery level, factors such as the battery consumption of the device over time and the possible battery monitoring errors caused by positioning drift are considered. For example, the quantization representation of the battery level in the spatio-temporal state drift vector is [5]. If the device temperature is also affected by spatio-temporal factors and positioning drift, and its quantization representation is [1] for example, then the obtained spatio-temporal state drift vector can be [4, 5, 1].

[0087] Finally, perform feature mapping on the spatio-temporal state drift vector to obtain a history-oriented positioning vector with adjustable limits. Feature mapping is an operation that converts the elements in the spatio-temporal state drift vector into a history-oriented positioning vector with adjustable limits according to a set mapping relationship. This mapping relationship is set based on the requirements of the entire system and the data processing logic. For example, the first element [4] in the spatio-temporal state drift vector may be converted into information related to the azimuth angle in the history-oriented positioning vector with adjustable limits through the feature mapping relationship. For example, after conversion according to the mapping rules, it is

[30] (the 30 in the embodiment of this application represents a certain value or range related to the azimuth angle, etc.). The second element [5] may be converted into distance-related information. For example, after conversion, it is

[12] (representing 12 nautical miles or other quantization representations related to distance). The third element [1] may be converted into information related to the stability of the device state. For example, after conversion, it is [low stability] (which can be a qualitative conversion result or a further quantization representation). Then, the obtained history-oriented positioning vector with adjustable limits can be [30, 12, low stability].

[0088] In this way, by combining the second orientation positioning drift tag with the first historical source state monitoring data to obtain the source state drift data, the positioning drift information can be incorporated into the source state monitoring, making the subsequent processed data more comprehensive. Identifying the spatio-temporal elements of the source state drift data yields a spatio-temporal state drift vector, further mining the spatio-temporal change information in the data. Finally, through feature mapping, a historical orientation positioning vector with adjustable limits is obtained. This method enables the historical orientation positioning vector to more accurately reflect the relevant information of the historical reporting device considering positioning drift, contributing to improving the utilization efficiency and processing accuracy of historical data in the entire technical solution, thereby enhancing the overall technical performance.

[0089] Under an exemplary technical idea, improving the model weights of the pre-debugged generative adversarial model based on the source state monitoring training data to obtain the target generative adversarial model includes: improving the model weights of the pre-debugged generative adversarial model based on the source state monitoring training data to obtain an intermediate generative adversarial model; performing perturbation feature detection on the historical source direction-finding interaction data to obtain a historical linked direction-finding interaction vector; the historical source direction-finding interaction data is used to describe the historical joint search and rescue direction-finding records where the historical reporting device exists; based on the historical spatio-temporal orientation positioning tag of the historical reporting device, determining a collaborative direction-finding cross vector matching the historical spatio-temporal orientation positioning tag from a set of collaborative direction-finding cross vectors; adjusting the dynamic direction-finding quantization result of the direction-finding information of the historical reporting device in the historical linked direction-finding interaction vector to the collaborative direction-finding cross vector matching the historical spatio-temporal orientation positioning tag to obtain an adjusted historical linked direction-finding interaction vector; performing feature mining and perturbation processing on the third orientation positioning drift tag to generate an initial azimuth drift tag vector; inputting the adjusted historical linked direction-finding interaction vector and the initial azimuth drift tag vector into the intermediate generative adversarial model for spatio-temporal azimuth attention enhancement to obtain an enhanced azimuth drift tag vector; identifying the enhanced azimuth drift tag vector to generate spatio-temporal orientation positioning training tags; improving the model weights of the intermediate generative adversarial model based on the spatio-temporal orientation positioning training tags to obtain the target generative adversarial model.

[0090] Furthermore, improving the model weights of the intermediate generative adversarial model based on the spatio-temporal orientation positioning training tags to obtain the target generative adversarial model includes: determining the comparison result between the spatio-temporal orientation positioning training tags and the historical spatio-temporal orientation positioning tags; improving the model weights of the intermediate generative adversarial model based on the comparison result between the spatio-temporal orientation positioning training tags and the historical spatio-temporal orientation positioning tags to obtain the target generative adversarial model.

[0091] In the process of determining the target generative adversarial model in the embodiments of the present application, first, the model weights of the pre-debugged generative adversarial model are improved based on the source state monitoring training data to obtain an intermediate generative adversarial model. The source state monitoring training data contains information related to source state monitoring, and this information is obtained through a series of previous operations. For example, the source state monitoring training data may contain the quantization values of the signal strength, power, and other state information of the historical reporting devices after processing, and these quantization values reflect the state characteristics of the devices in different situations. By using these data to improve the weights of the pre-debugged generative adversarial model, the model begins to adjust in a direction more consistent with the actual data characteristics, thereby obtaining an intermediate generative adversarial model.

[0092] Next, perturbation feature detection is performed on the historical source direction finding interaction data to obtain the historical linkage direction finding interaction vector. The historical source direction finding interaction data is used to describe the historical joint search and rescue direction finding records of the historical reporting devices. The direction finding information of the historical reporting devices in the historical source direction finding interaction data contains key direction finding information such as azimuth and distance. For example, the azimuth of the historical reporting device is 45°, and the distance is 12 nautical miles. When performing perturbation feature detection, due to the influence of environmental factors (such as magnetic field interference, meteorological conditions, etc.), these direction finding information may be perturbed. For example, the perturbation degree of the azimuth is quantified as 1 (indicating slight perturbation), and the perturbation degree of the distance is quantified as 2 (indicating medium perturbation). Then, the obtained historical linkage direction finding interaction vector may contain the quantization values of these direction finding information and their perturbation degrees, such as [45, 1, 12, 2] (this is only a simple representation in the embodiments of the present application, and actually may contain more information).

[0093] Then, based on the historical spatio-temporal orientation and positioning label of the historical reporting device, a collaborative direction finding cross vector matching the historical spatio-temporal orientation and positioning label is determined from the collaborative direction finding cross vector set. The historical spatio-temporal orientation and positioning label is a positioning identifier of the historical reporting device in the spatio-temporal dimension, and contains multi-faceted information such as azimuth, distance, signal strength, and power. The collaborative direction finding cross vector set is a previously constructed set containing multiple collaborative direction finding cross vectors, and each collaborative direction finding cross vector contains the integrated result of the direction finding information and state information related to the reporting device. For example, the historical spatio-temporal orientation and positioning label is [40, 10, 3, 4] (representing the quantization values of azimuth, distance, signal strength, and power respectively), and a collaborative direction finding cross vector matching it is searched in the collaborative direction finding cross vector set. For example, the found collaborative direction finding cross vector is [40, 3, 10, 4] (the order and numerical values in the embodiments of the present application are obtained according to the matching rules).

[0094] After that, adjust the dynamic direction-finding quantization result of the direction-finding information of the historical reporting device in the historical linkage direction-finding interaction vector to the collaborative direction-finding cross vector matching the historical space-time orientation positioning label, and obtain the adjusted historical linkage direction-finding interaction vector. Taking the previous historical linkage direction-finding interaction vector [45, 1, 12, 2] as an example, adjust the azimuth 45 and the distance 12 in it to the values in the matching collaborative direction-finding cross vector, that is, obtain the adjusted historical linkage direction-finding interaction vector [40, 3, 10, 2] (actual operation may involve adjustments of more elements).

[0095] Then, perform feature mining and perturbation processing on the third orientation positioning drift label to generate an initial bearing drift label vector. The third orientation positioning drift label contains information related to positioning drift. Through feature mining and perturbation processing, this information is transformed into an initial bearing drift label vector. For example, the third orientation positioning drift label contains information such as the possible drift trend of positioning in a certain direction. After processing, the initial bearing drift label vector may contain a quantitative representation of this drift trend, such as [2, 1] (the numerical values in the embodiments of this application represent drift quantization in different dimensions).

[0096] Subsequently, input the adjusted historical linkage direction-finding interaction vector and the initial bearing drift label vector into the intermediate generative adversarial model for spatio-temporal bearing attention enhancement, and obtain the enhanced bearing drift label vector. The intermediate generative adversarial model processes the input vectors according to its internal structure and algorithm, and enhances the information related to spatio-temporal bearing attention. For example, the input adjusted historical linkage direction-finding interaction vector is [40, 3, 10, 2], and the initial bearing drift label vector is [2, 1]. After being processed by the model, the enhanced bearing drift label vector obtained is, for example, [42, 4, 11, 3] (the numerical values in the embodiments of this application are the results after being processed by the model, reflecting the characteristics after spatio-temporal bearing attention enhancement).

[0097] Next, identify the enhanced bearing drift label vector to generate a spatio-temporal orientation positioning training label. The identification process interprets the elements in the enhanced bearing drift label vector according to the preset rules. For example, if the first element in the enhanced bearing drift label vector represents a certain quantization result of the azimuth, the second element represents the quantization result of the distance, the third element represents the quantization result related to the signal strength, and the fourth element represents the quantization result related to the battery power, then the spatio-temporal orientation positioning training label generated according to these elements can be "the angle represented by the numerical value 42 corresponding to the azimuth, the distance represented by the numerical value 4 corresponding to the distance, the strength represented by the numerical value 11 corresponding to the signal strength, and the battery power state represented by the numerical value 3 corresponding to the battery power".

[0098] Finally, determine the comparison result between the spatio-temporal orientation and positioning training label and the historical spatio-temporal orientation and positioning label, and improve the model weights of the intermediate generative adversarial model based on this comparison result to obtain the target generative adversarial model. For example, the spatio-temporal orientation and positioning training label is [42, 4, 11, 3], and the historical spatio-temporal orientation and positioning label is [40, 10, 3, 4]. By comparing the differences between each element, such as the azimuth angle changing from 40 to 42 and the distance changing from 10 to 4, etc. According to these differences, improve the weights of the intermediate generative adversarial model according to certain rules (such as weighted average, etc.), so that the model can more accurately predict and process relevant information, and finally obtain the target generative adversarial model.

[0099] Thus, through multiple rounds of adjustment and improvement, first obtain the intermediate generative adversarial model based on the source state monitoring training data, and then through a series of operations including the processing of historical data, vector adjustment, model input and output processing, and finally comparison and improvement, the model can make full use of the information in the historical data. By continuously adjusting the model weights, the processing ability of the model for spatio-temporal orientation and positioning related information is improved, making the model more adaptable to the actual situation, thereby enhancing the accuracy and reliability of the entire technical solution when processing tasks related to the target warning device, and being able to perform operations such as more accurately positioning the target.

[0100] To facilitate the understanding of the above technical solution, another application scenario is used to introduce the embodiments of the present application. For example, there are 5 search and rescue platforms A, B, C, D, and E, performing a search and rescue mission for a wrecked ship (target warning device) in a vast sea area.

[0101] First, each platform will obtain source direction-finding interaction data regarding the target warning device, and these data are used to describe the joint search and rescue direction-finding records of the area where the target warning device exists.

[0102] For platform A, the direction-finding information obtained by its first reporting device shows that the azimuth angle of the target reporting device relative to platform A is 30° and the distance is 10 nautical miles. The azimuth angle obtained by platform B is 40° and the distance is 12 nautical miles; the azimuth angle of platform C is 50° and the distance is 15 nautical miles; the azimuth angle of platform D is 35° and the distance is 13 nautical miles; the azimuth angle of platform E is 45° and the distance is 14 nautical miles. The direction-finding information of the first reporting devices of these respective platforms constitutes the source direction-finding interaction data. Then, feature mining is performed on this source direction-finding interaction data to obtain the linked direction-finding interaction vector. For example, for the azimuth angle, with a quantization interval of 15°, the 30° of platform A is in the second interval and is quantified as 2; for the distance, with a quantization interval of 5 nautical miles, 10 nautical miles is in the second interval and is quantified as 2. Quantify the other platforms in the same way. The azimuth angle of platform B is quantified as 3 and the distance is quantified as 3; the azimuth angle of platform C is quantified as 4 and the distance is quantified as 4; the azimuth angle of platform D is quantified as 2 and the distance is quantified as 3; the azimuth angle of platform E is quantified as 3 and the distance is quantified as 3. In this way, the linked direction-finding interaction vector obtained contains the dynamic direction-finding quantization results of the direction-finding information of the first reporting devices of each platform, such as [2, 2, 3, 3, 4, 4, 2, 3, 3, 3] (corresponding to the azimuth angle and distance quantization values of platforms A - E respectively).

[0103] Next, obtain the target multi-platform collaborative direction-finding code. The first source state monitoring data represents the real-time monitoring state of the target reporting device. For example, the signal strength of the target reporting device is medium (quantified as 1), and the battery level is 80% (quantified as 4 according to an interval of every 20%). Integrate the state monitoring description vector of this first source state monitoring data with the dynamic direction-finding quantization result of the direction-finding information of the first reporting device to obtain the target multi-platform collaborative direction-finding code. For example, for platform A, the integrated code part can be [2, 2, 1, 4], for platform B it is [3, 3, 1, 4], and so on. Finally, the target multi-platform collaborative direction-finding code is [2, 2, 1, 4, 3, 3, 1, 4, 4, 4, 1, 4, 2, 3, 1, 4, 3, 3, 1, (corresponding to the direction-finding quantization results and state monitoring description vectors of platforms A - E respectively).

[0104] Then, based on the dynamic direction-finding quantization result of the direction-finding information of the first reporting device in the linked direction-finding interaction vector and the target multi-platform collaborative direction-finding coding, the target linked direction-finding interaction vector is determined. A possible way is based on a weighting algorithm. For example, the weight of the azimuth quantization value part is 0.6, the weight of the distance quantization value part is 0.3, the signal strength weight is 0.05, and the power weight is 0.05. Taking platform A as an example, its azimuth quantization value is 2, the distance quantization value is 2, the signal strength quantization value is 1, and the power quantization value is 4. Calculated according to the weighting algorithm: for the azimuth part, it is 2×0.6 = 1.2; for the distance part, it is 2×0.3 = 0.6; for the signal strength part, it is 1×0.05 = 0.05; for the power part, it is 4×0.05 = 0.2. Combining these results in a certain order, a new sub-vector is obtained for platform A, such as [1.2, 0.6, 0.05, 0.2]. Using the same method to calculate for platforms B - E, and then combining these sub-vectors to obtain the target linked direction-finding interaction vector, such as [1.2, 0.6, 0.05, 0.2, 1.8, 0.9, 0.05, 0.2, 2.4, 1.2, 0.05, 0.2, 1.4, 0.9, 0.05, 0.2, 1.8, 0.9, 0.05, 0.2].

[0105] Finally, based on the target linked direction-finding interaction vector, a spatio-temporal orientation and positioning label corresponding to the regional joint search and rescue direction-finding record is generated. This spatio-temporal orientation and positioning label can accurately identify the position and status of the target reporting device in the spatio-temporal dimension. For example, according to the values in the target linked direction-finding interaction vector, through calculation and interpretation, it can be obtained that at a certain moment (such as 10:00 on July 1, 2023), the target reporting device is located at a specific longitude and latitude position (such as 120° east longitude, 30° north latitude), and its source state (such as medium signal strength, 80% power) is also included in this label. The spatio-temporal orientation and positioning label can be "2023-07-01 10:00, azimuth (calculated according to the vector), distance (calculated according to the vector), 120° east longitude, 30° north latitude, medium signal strength, 80% power". Through the above spatio-temporal orientation and positioning label, the 5 search and rescue platforms can more accurately locate the target reporting device, thereby improving the efficiency of maritime formation search and rescue.

[0106] As an optional but non-limiting embodiment, after generating the spatio-temporal orientation and positioning label corresponding to the regional joint search and rescue direction-finding record based on the target linked direction-finding interaction vector, it further includes:

[0107] Step 150a: Obtain the historical spatio-temporal orientation and positioning tag set corresponding to the regional joint search and rescue direction finding record; perform spatio-temporal azimuth offset comparison between the spatio-temporal orientation and positioning tag and the historical spatio-temporal orientation and positioning tag set to obtain the target spatio-temporal azimuth offset; perform azimuth compensation adjustment on the spatio-temporal orientation and positioning tag based on the target spatio-temporal azimuth offset to generate a compensated target spatio-temporal orientation and positioning tag; extract the multi-platform collaborative direction finding path features in the compensated target spatio-temporal orientation and positioning tag, and generate an optimized path planning vector based on the multi-platform collaborative direction finding path features; perform spatio-temporal trajectory superposition between the optimized path planning vector and the compensated target spatio-temporal orientation and positioning tag to generate the optimized spatio-temporal orientation and positioning tag corresponding to the regional joint search and rescue direction finding record; wherein, the optimized path planning vector is used to describe the dynamic collaborative direction finding path of the target notification device in the regional joint search and rescue direction finding record.

[0108] In step 150a, obtain the historical spatio-temporal orientation and positioning tag set corresponding to the regional joint search and rescue direction finding record. The historical spatio-temporal orientation and positioning tag set contains the positioning identification data of the target notification device generated in past search and rescue missions in this sea area. For example, during three joint search and rescue missions executed in June 2023, the spatio-temporal orientation and positioning tag sets generated by the platform cluster for the emergency beacon of the same shipwreck. Compare the currently generated spatio-temporal orientation and positioning tag with this tag set for spatio-temporal azimuth offset. For example, the current tag indicates that the target notification device is located at 121.5 degrees east longitude and 30.2 degrees north latitude, while the coordinates of the three most recent records in the historical tag set are 121.3 degrees east longitude and 30.1 degrees north latitude, 121.4 degrees east longitude and 30.15 degrees north latitude, and 121.45 degrees east longitude and 30.18 degrees north latitude. By calculating the moving average and standard deviation of the current coordinates and the historical coordinate sequence, the target spatio-temporal azimuth offset is obtained as a drift trend of 0.2 degrees per hour in the northeast direction. Perform azimuth compensation adjustment on the current tag based on this offset. For example, add an inverse drift correction amount to the original coordinates to generate a compensated target spatio-temporal orientation and positioning tag, and correct the coordinates to 121.48 degrees east longitude and 30.19 degrees north latitude. Extract the collaborative movement trajectory features of the platform cluster in this compensated tag. For example, the encircling path of platform A from southwest to northeast and platform B from southeast to northwest. Generate an optimized path planning vector based on these multi-platform collaborative direction finding path features. This vector encodes the azimuth change rate and distance convergence rate of the best approximation path of each platform. Perform spatio-temporal trajectory superposition between this vector and the compensated tag. For example, integrate the optimized path parameters of platform C into the coordinate prediction model, and finally generate an optimized spatio-temporal orientation and positioning tag containing the dynamic collaborative direction finding path. This tag not only contains the corrected geographical location but also embeds the optimal navigation path parameters for each search and rescue platform to reach the target.

[0109] As an optional but non-limiting embodiment, after generating the spatio-temporal orientation and positioning tag corresponding to the regional joint search and rescue direction finding record based on the target linkage direction finding interaction vector, the following steps are further included:

[0110] Step 150b: Perform dynamic drift monitoring on the spatio-temporal orientation and positioning tag, and extract the positioning drift monitoring data corresponding to the spatio-temporal orientation and positioning tag; generate a dynamic drift compensation vector according to the positioning drift monitoring data, and perform spatio-temporal trajectory matching between the dynamic drift compensation vector and the spatio-temporal orientation and positioning tag; correct the drift characteristics of the spatio-temporal orientation and positioning tag based on the matching result to obtain a corrected spatio-temporal orientation and positioning tag; obtain the real-time direction finding cooperation data of multiple platforms in the regional joint search and rescue direction finding record, and perform spatio-temporal azimuth synchronization between the real-time direction finding cooperation data and the corrected spatio-temporal orientation and positioning tag; adjust the dynamic direction finding quantization result in the target linkage direction finding interaction vector according to the synchronized result to generate an updated target linkage direction finding interaction vector; regenerate the spatio-temporal orientation and positioning tag corresponding to the regional joint search and rescue direction finding record based on the updated target linkage direction finding interaction vector.

[0111] In step 150b, dynamic drift monitoring is performed on the spatio-temporal orientation and positioning tag. In a maritime search and rescue scenario, the target notification device may be affected by ocean currents and cause position drift. For example, the static coordinates indicated by the tag cannot reflect the real-time position change. By extracting the continuous change data of the azimuth angle parameter in the tag, it is monitored that the target has a northeastward offset of 0.05 degrees / minute in the last 15 minutes. A dynamic drift compensation vector is generated according to this positioning drift monitoring data, and this vector includes a longitude compensation coefficient of 1.002 and a latitude compensation coefficient of 1.0015. Perform spatio-temporal trajectory matching between this vector and the original spatio-temporal orientation and positioning tag. For example, apply the compensation coefficient to the coordinate transformation matrix to make the predicted trajectory coincide with the measured drift pattern. Correct the drift characteristics of the tag based on the matching result. For example, on the basis of the original coordinates of 121.5 degrees east longitude and 30.2 degrees north latitude, superimpose dynamic compensation to generate a corrected spatio-temporal orientation and positioning tag of 121.503 degrees east longitude and 30.2018 degrees north latitude. Obtain the real-time direction finding cooperation data of the platform cluster. For example, the azimuth angle of 42.5 degrees and the distance of 13.2 nautical miles measured by platform D are newly uploaded. Perform spatio-temporal azimuth synchronization between this data and the corrected tag, and verify that the error between the coordinate predicted value and the measured value is within 0.1 nautical mile. Adjust the dynamic direction finding quantization result in the target linkage direction finding interaction vector according to the synchronization result. For example, update the azimuth angle quantization value of platform D from 3 to 4, and adjust the distance quantization value from 3 to 3.2 to generate an updated target linkage direction finding interaction vector. Regenerate the spatio-temporal orientation and positioning tag based on this updated vector, and output the coordinates of 121.505 degrees east longitude and 30.2023 degrees north latitude with real-time drift compensation and the latest device status parameters.

[0112] As an optional but non-limiting embodiment, after generating the spatio-temporal orientation and positioning tag corresponding to the regional joint search and rescue direction finding record based on the target linkage direction finding interaction vector, the following steps are further included:

[0113] Step 150c: Obtain supplementary direction finding interaction data of other collaborative search and rescue platforms in the regional joint search and rescue direction finding record; extract spatio-temporal elements from the supplementary direction finding interaction data to generate a supplementary direction finding spatio-temporal feature vector; perform multi-source direction finding feature fusion on the supplementary direction finding spatio-temporal feature vector and the spatio-temporal orientation and positioning tag to generate a fused direction finding trajectory vector; calibrate the trajectory of the spatio-temporal orientation and positioning tag according to the fused direction finding trajectory vector to generate a calibrated spatio-temporal orientation and positioning tag; extract dynamic collaborative direction finding parameters from the calibrated spatio-temporal orientation and positioning tag, and update the target multi-platform collaborative direction finding code based on the dynamic collaborative direction finding parameters; re-integrate the updated target multi-platform collaborative direction finding code with the target linkage direction finding interaction vector to generate a fused spatio-temporal orientation and positioning tag corresponding to the regional joint search and rescue direction finding record.

[0114] In step 150c, obtain supplementary direction finding interaction data of other collaborative search and rescue platforms in the regional joint search and rescue direction finding record. For example, the direction finding data transmitted back by the newly deployed aerial search and rescue platform F, including high-precision measurement values of the azimuth angle of 38.7 degrees and the distance of 9.8 nautical miles. Extract spatio-temporal elements from this supplementary data to generate a supplementary direction finding spatio-temporal feature vector including the timestamp 2023-07-01T11:05:00 and the three-dimensional coordinate offset of 0.03 degrees. Perform multi-source direction finding feature fusion on this feature vector and the existing spatio-temporal orientation and positioning tag. For example, fuse the data of platform F with the measurement values of platforms A - E through the Kalman filter algorithm to generate a fused direction finding trajectory vector. This vector encodes the motion trajectory curvature parameter of 0.12 radians / nm for the target in the 11:00 - 11:05 period. Calibrate the original tag according to this vector. For example, correct the straight-line motion model to a curve motion model to generate a calibrated spatio-temporal orientation and positioning tag, and update the coordinates to 121.508 degrees east longitude and 30.2035 degrees north latitude. Extract the dynamic collaborative direction finding parameters from this calibrated tag, including the weight distribution of the direction finding data of each platform and the trajectory confidence index, and update the direction finding quantization value and status monitoring parameters of platform F in the target multi-platform collaborative direction finding code based on this. Re-integrate the updated collaborative direction finding code with the target linkage direction finding interaction vector. For example, append the azimuth angle quantization value of 3.8 and the distance quantization value of 2.9 of platform F at the end of the vector to generate a fused spatio-temporal orientation and positioning tag. This tag not only integrates the measurement data of the newly added platform but also improves the positioning accuracy through feature fusion, and finally outputs the optimized coordinates of 121.509 degrees east longitude and 30.2041 degrees north latitude and the trajectory confidence level of 0.95 including the collaborative data of six platforms.

[0115] In summary, in the embodiment of the present application, by performing feature mining on the source direction-finding interaction data to obtain the linkage direction-finding interaction vector, the original direction-finding information is converted into a dynamic direction-finding quantization result, which can effectively integrate the direction-finding information of multiple platforms and make it more valuable for processing. Obtaining the target multi-platform collaborative direction-finding coding integrates the direction-finding result and the real-time status monitoring data of the target reporting device, making the positioning information more comprehensive. Determining the target linkage direction-finding interaction vector further fuses multi-source information, improving the accuracy and reliability of the information. Generating the spatio-temporal orientation positioning label based on this vector can accurately determine the position and status of the target reporting device in the spatio-temporal dimension, which greatly improves the search and rescue efficiency in the regional joint search and rescue scenario, enhances the accuracy and credibility of target positioning, and reduces problems such as search and rescue delays caused by inaccurate positioning.

[0116] Furthermore, Figure 2 FIG. 6 is a schematic structural diagram of a dynamic limit orientation positioning system 200 provided by an embodiment of the present application. As Figure 2 shown, the dynamic limit orientation positioning system 200 includes a processor 210. The processor 210 can call and run a computer program from a memory to implement the method in the embodiment of the present application. Optionally, as Figure 2 shown, the dynamic limit orientation positioning system 200 may further include a memory 230. Among them, the processor 210 can call and run a computer program from the memory 230 to implement the method in the embodiment of the present application. Among them, the memory 230 can be a separate device independent of the processor 210, or can be integrated in the processor 210. Optionally, as Figure 2 shown, the dynamic limit orientation positioning system 200 may further include a transceiver 220. The processor 210 can control the transceiver 220 to interact with other devices. Specifically, it can send information or data to other devices, or receive information or data sent by other devices. Optionally, the dynamic limit orientation positioning system 200 can implement the corresponding processes of the storage engine or components in the storage engine (such as a processing module) or the device deploying the storage engine in each method of the embodiment of the present application. For the sake of brevity, it will not be elaborated here. It should be understood that the processor in the embodiment of the present application may be an integrated circuit chip with signal processing capabilities. It can be understood that the memory in the embodiment of the present application can be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. It should be noted that the memory of the system and method described herein is intended to include, but is not limited to, suitable types of memories.

[0117] Based on the above, a readable storage medium is provided. A program or instruction is stored on the readable storage medium, and when the program or instruction is executed by a processor, the steps of the above method are implemented.

[0118] The embodiments of the present application have been described above in conjunction with the accompanying drawings. However, the embodiments of the present application are not limited to the above specific implementation manners. The above specific implementation manners are merely illustrative rather than restrictive. Under the inspiration of the embodiments of the present application, those of ordinary skill in the art can also make many forms without departing from the purpose of the embodiments of the present application and the scope protected by the embodiments of the present application, and all of them fall within the protection scope of the embodiments of the present application.

Claims

1. A dynamic limit directional positioning method applied to joint search and rescue of multiple platforms at sea, characterized in that, The method is implemented through a dynamic limit directional positioning system, and the method includes: Performing feature mining on the source direction finding interaction data to obtain a linkage direction finding interaction vector; the source direction finding interaction data is used to describe the regional joint search and rescue direction finding record where the target notification device exists, and the source direction finding interaction data includes the first notification device direction finding information characterizing the target notification device; the dynamic direction finding quantization result of the first notification device direction finding information is included in the linkage direction finding interaction vector; Obtaining a target multi-platform collaborative direction finding code; the target multi-platform collaborative direction finding code is obtained by integrating the state monitoring description vector of the first source state monitoring data on the basis of the dynamic direction finding quantization result of the first notification device direction finding information, and the first source state monitoring data characterizes the real-time monitoring state of the target notification device; Determining a target linkage direction finding interaction vector through the dynamic direction finding quantization result of the first notification device direction finding information in the linkage direction finding interaction vector and the target multi-platform collaborative direction finding code; Generating a spatio-temporal directional positioning label corresponding to the regional joint search and rescue direction finding record according to the target linkage direction finding interaction vector.

2. The method according to claim 1, wherein The obtaining of the target multi-platform collaborative direction finding code includes: Determining a platform collaborative direction finding knowledge graph in the platform database; the platform collaborative direction finding knowledge graph includes platform collaborative direction finding knowledge corresponding to multiple second notification device direction finding information respectively; Screening the platform collaborative direction finding knowledge corresponding to the first notification device direction finding information from the platform collaborative direction finding knowledge graph to obtain the target multi-platform collaborative direction finding code; The steps of generating the platform collaborative direction finding knowledge corresponding to the second notification device direction finding information include: Obtaining multiple second source state monitoring data of the notification device characterized by the second notification device direction finding information; Performing spatio-temporal element identification on the multiple second source state monitoring data respectively to obtain multiple spatio-temporal element state vectors, and performing perturbation feature detection on the second notification device direction finding information to obtain the perturbation state detection vector of the target notification device; Performing feature integration according to the perturbation state detection vector and the multiple spatio-temporal element state vectors to obtain the platform collaborative direction finding knowledge corresponding to the second notification device direction finding information; The performing feature integration according to the perturbation state detection vector and the multiple spatio-temporal element state vectors to obtain the platform collaborative direction finding knowledge corresponding to the second notification device direction finding information includes: For each spatio-temporal element state vector in the multiple spatio-temporal element state vectors, integrating the spatio-temporal element state vector on the basis of the perturbation state detection vector to obtain a collaborative direction finding cross vector; Processing each of the collaborative direction finding cross vectors to obtain the platform collaborative direction finding knowledge of the target notification device.

3. The method according to claim 1, wherein The obtaining of the target multi-platform collaborative direction finding code includes: Performing spatio-temporal element identification on the second source state monitoring data of the target notification device to obtain the spatio-temporal element state vector of the target notification device; Performing perturbation feature detection on the first notification device direction finding information to obtain the perturbation state detection vector of the target notification device; Integrate the spatio-temporal element state vector of the target notification device based on the disturbance state detection vector of the target notification device to obtain the target multi-platform collaborative direction finding code.

4. The method according to any one of claims 1 to 3, characterized in that, The generating the spatio-temporal orientation and positioning label corresponding to the regional joint search and rescue direction finding record according to the target linkage direction finding interaction vector includes: Generating an orientation and positioning vector with adjustable limit according to the first orientation and positioning drift label; Inputting the target linkage direction finding interaction vector and the orientation and positioning vector with adjustable limit into a target generative adversarial model for spatio-temporal phase attention enhancement to obtain a spatio-temporal phase attention vector; Identifying the spatio-temporal phase attention vector to generate the spatio-temporal orientation and positioning label corresponding to the regional joint search and rescue direction finding record; Among them, the target generative adversarial model is obtained by improving a pre-debugged generative adversarial model based on the collaborative direction finding cross vector of historical notification devices. The collaborative direction finding cross vector of historical notification devices is the result of feature integration of the second source state monitoring data of the historical notification device and the direction finding information of the second notification device of the historical notification device.

5. The method according to claim 4, wherein The steps of obtaining the target generative adversarial model include: Performing spatio-temporal element identification and disturbance processing on the first historical source state monitoring data to obtain an orientation and positioning vector with adjustable limit; Integrating the dynamic direction finding quantization result of the historical notification device direction finding information with the spatio-temporal element state vector of the second historical source state monitoring data to obtain a historical collaborative direction finding cross vector; the first historical source state monitoring data and the second historical source state monitoring data are the source state monitoring data of the same historical notification device, and the historical notification device direction finding information is the direction finding interaction data representing the historical notification device; Inputting the orientation and positioning vector with adjustable limit and the historical collaborative direction finding cross vector into the pre-debugged generative adversarial model for spatio-temporal phase attention enhancement to obtain a target spatio-temporal element state vector; Identifying the target spatio-temporal element state vector to generate source state monitoring training data; Improving the model weights of the pre-debugged generative adversarial model according to the source state monitoring training data to obtain a target generative adversarial model.

6. The method according to claim 5, wherein The integrating the dynamic direction finding quantization result of the historical notification device direction finding information with the spatio-temporal element state vector of the second historical source state monitoring data to obtain a historical collaborative direction finding cross vector includes: Performing disturbance feature detection on the historical notification device direction finding information to obtain a historical disturbance state detection vector; Performing spatio-temporal element identification on the second historical source state monitoring data to obtain a historical spatio-temporal element state vector; Performing feature integration on the historical disturbance state detection vector and the historical spatio-temporal element state vector to obtain a historical collaborative direction finding cross vector.

7. The method according to claim 5, wherein The first historical source state monitoring data and the second historical source state monitoring data are different; The improving the model weights of the pre-debugged generative adversarial model according to the source state monitoring training data to obtain a target generative adversarial model includes: Calculate a first training error based on a comparison result between the source state monitoring training data and the second historical source state monitoring data; Improve the model weights of the pre-debugged generative adversarial model using the first training error to obtain a target generative adversarial model.

8. The method according to claim 5, characterized in that, The obtaining of the adjustable historical orientation and positioning vector by performing spatio-temporal element identification and perturbation processing on the first historical source state monitoring data includes: Combine the second orientation and positioning drift label with the first historical source state monitoring data to obtain source state drift data; Perform spatio-temporal element identification on the source state drift data to obtain a spatio-temporal state drift vector; Perform feature mapping on the spatio-temporal state drift vector to obtain an adjustable historical orientation and positioning vector.

9. The method according to claim 5, wherein The obtaining of the target generative adversarial model by improving the model weights of the pre-debugged generative adversarial model based on the source state monitoring training data includes: Improve the model weights of the pre-debugged generative adversarial model based on the source state monitoring training data to obtain an intermediate generative adversarial model; Perform perturbation feature detection on historical source direction finding interaction data to obtain a historical linked direction finding interaction vector; the historical source direction finding interaction data is used to describe the historical joint search and rescue direction finding record of the historical notification device; Based on the historical spatio-temporal orientation and positioning label of the historical notification device, determine a collaborative direction finding cross vector that matches the historical spatio-temporal orientation and positioning label from a set of collaborative direction finding cross vectors; Adjust the dynamic direction finding quantization result of the direction finding information of the historical notification device in the historical linked direction finding interaction vector to the collaborative direction finding cross vector that matches the historical spatio-temporal orientation and positioning label to obtain an adjusted historical linked direction finding interaction vector; Perform feature mining and perturbation processing on the third orientation and positioning drift label to generate an initial azimuth drift label vector; Input the adjusted historical linked direction finding interaction vector and the initial azimuth drift label vector into the intermediate generative adversarial model for spatio-temporal azimuth attention enhancement to obtain an enhanced azimuth drift label vector; Perform identification on the enhanced azimuth drift label vector to generate spatio-temporal orientation and positioning training labels; Improve the model weights of the intermediate generative adversarial model based on the spatio-temporal orientation and positioning training labels to obtain a target generative adversarial model; The obtaining of the target generative adversarial model by improving the model weights of the intermediate generative adversarial model based on the spatio-temporal orientation and positioning training labels includes: Determine a comparison result between the spatio-temporal orientation and positioning training labels and the historical spatio-temporal orientation and positioning labels; Improve the model weights of the intermediate generative adversarial model based on the comparison result between the spatio-temporal orientation and positioning training labels and the historical spatio-temporal orientation and positioning labels to obtain a target generative adversarial model.

10. A dynamic limit directional positioning system, characterized in that, Includes at least one processor and a memory; the memory stores computer execution instructions; the at least one processor executes the computer execution instructions stored in the memory, such that the at least one processor executes the method according to any one of claims 1-9.

Images