A method, device and medium for assisting decision-making in searching for distressed targets at sea.

Abstract

This application provides a method, device, and medium for assisting decision-making in the search of distressed targets at sea, relating to the field of maritime technology. The method includes: acquiring marine meteorological and environmental forecast information for the distressed area and surrounding waters to determine the search area; calculating the sea area risk level and the urgency of the search operation in the search area; acquiring search resources; establishing a search resource decision-making model, combining the urgency of the search operation, the sea area risk level, and the search resources to obtain the optimal search resource planning scheme; dividing the search area into sub-regions according to the sea area risk level, and combining the optimal search resource planning scheme to obtain the working sea area of ​​the search forces and determine the optimal global path planning for the search forces. This scheme achieves a balance between search speed, search cost, and search safety in search operations, improving the efficiency of maritime search operations.

Description

Technical Field

[0001] This application relates to the field of maritime technology, and in particular to a method, device and medium for assisting decision-making in the search of distressed targets at sea. Background Technology

[0002] With the rapid development of the marine economy, the frequency of maritime accidents has also increased, making search and rescue operations for maritime distress a crucial part of maritime safety. Generally, the goals of search operation planning are to minimize search time, reduce search resource costs, and maximize search safety, while ensuring successful rescue. The emphasis of search plan evaluation should differ depending on the urgency of the maritime distress. However, existing plans typically only consider search time or search resource costs when planning maritime search operations, resulting in a relatively singular focus. Furthermore, a quantitative evaluation method is lacking when comparing different search plans. In conclusion, the planning and resource allocation for maritime accident search operations should consider more factors, and existing plans cannot provide a comprehensive plan for maritime search operations. Summary of the Invention

[0003] The purpose of this invention is to provide a method, device, and medium for assisting decision-making in the search for distressed targets at sea, in order to solve the problem that existing maritime search schemes cannot comprehensively plan maritime search operations.

[0004] The above-mentioned objective of this application is achieved through the following technical solution:

[0005] S1: Obtain marine meteorological and environmental forecasts for the distress area and surrounding waters to determine the search area;

[0006] S2: Calculate the risk level of the search area and the urgency of the search operation;

[0007] S3: Acquire search resources; establish a search resource decision model, and combine the urgency of the search operation, the risk level of the sea area, and the search resources to obtain the best search resource planning scheme based on search time, search cost, and search safety.

[0008] S4: Based on the risk level of the sea area, the search area is divided into sub-regions. Combined with the best search resource planning scheme, the working sea area of ​​the search force is obtained and the optimal global path planning of the search force is determined.

[0009] An electronic device includes a processor, a memory, a user interface, and a network interface. The memory is used to store instructions, the user interface and the network interface are used to communicate with other devices, and the processor is used to execute the instructions stored in the memory to enable the electronic device to perform a method, device, and medium for assisting decision-making in searching for distressed targets at sea.

[0010] A computer-readable storage medium storing instructions that, when executed, perform a method, apparatus, and medium for assisting decision-making in searching for distressed targets at sea.

[0011] The beneficial effects of the technical solution provided in this application are:

[0012] 1. Taking into account the risks of the search and rescue sea area and the urgency of the search and rescue operation, a search resource decision-making scheme was rationally planned based on search time, search cost, and search safety. A specific algorithm was introduced to quickly allocate search and rescue resources, adapt to the dynamically changing environment, and a comprehensive search action planning system was constructed, which improved the efficiency of maritime search action planning and resource utilization.

[0013] 2. This invention provides a method, device, and medium for assisting decision-making in the search for distressed targets at sea. The method includes a search operation risk and urgency assessment module, a search resource planning module, and a search route planning module, providing comprehensive planning solutions for search operations in maritime distress incidents. Specifically, the search operation urgency module uses variables such as the number of people in distress, the type of distress incident, and sea temperature, employing empirical assessment methods and the analytic hierarchy process (AHP) to quantify the urgency level. The search resource planning module aims to minimize search operation time, minimize search resource costs, and maximize search operation safety. It establishes a search resource planning model, adjusts the weight allocation of search time, search cost, and search safety in the search plan planning model based on the urgency level, and uses index evaluation normalization processing to evaluate the search plan. The search route planning module includes a module for determining the overall search area and a global search method decision module. Attached Figure Description

[0014] The present application will be further described below with reference to the accompanying drawings and embodiments. In the accompanying drawings:

[0015] Figure 1 This is a step diagram of an embodiment of this application;

[0016] Figure 2 This is a schematic diagram of the electronic device structure in the embodiments of this application. Detailed Implementation

[0017] To provide a clearer understanding of the technical features, objectives, and effects of this application, the specific embodiments of this application will now be described in detail with reference to the accompanying drawings.

[0018] Embodiments of this application provide a method, device, and medium for assisting decision-making in searching for distressed targets at sea.

[0019] Please refer to Figure 1 , Figure 1This is a step diagram of a method, device, and medium for assisting decision-making in searching for distressed targets at sea, as described in an embodiment of this application, including:

[0020] S1: Obtain marine meteorological and environmental forecasts for the distress area and surrounding waters to determine the search area;

[0021] S2: Calculate the risk level of the search area and the urgency of the search operation;

[0022] S3: Acquire search resources; establish a search resource decision model, and combine the urgency of the search operation, the risk level of the sea area, and the search resources to obtain the best search resource planning scheme based on search time, search cost, and search safety.

[0023] S4: Based on the risk level of the sea area, the search area is divided into sub-regions. Combined with the best search resource planning scheme, the working sea area of ​​the search force is obtained and the optimal global path planning of the search force is determined.

[0024] Step S1 includes:

[0025] S11: Establish a wind-wave-current coupled numerical forecasting model and obtain marine meteorological environment forecast information for the distress area and nearby sea areas through the wind-wave-current coupled numerical forecasting model;

[0026] The wind-wave-current coupled numerical prediction model includes: an atmospheric module, a wave module, and an ocean current module;

[0027] As one embodiment, the wind-wave-current coupled numerical forecasting model consists of an atmospheric module, a wave module, and an ocean current module. This model comprehensively considers the atmospheric, wave, and ocean current modules to simulate complex dynamic processes in the marine environment. The atmospheric module predicts the wind field (wind speed and direction), providing the ocean surface dynamics source and driving the evolution of waves and currents; the wave module calculates the generation, propagation, development, and breaking processes of waves, and couples with the wind field and ocean currents; the ocean current module simulates the tides and ocean current fields, considering ocean current variations caused by factors such as wind, temperature, and salinity. Based on WRF (Weather Research and Forecasting)-FVCOM (Finite-Volume Coastal Ocean Model)-SWAVE (Simulating Waves), the wind-wave-current coupled numerical forecasting model can be further optimized using a deep learning correction model.

[0028] As one example, the WRF-FVCOM-SWAVE coupled model is a coupled system integrating meteorological, oceanographic, and wave models. WRF (Weather Research and Forecasting Model) is a high-resolution numerical weather prediction model primarily used for weather forecasting and climate research. It can simulate various physical processes in the atmosphere, such as temperature, humidity, and wind speed. FVCOM (Finite Volume Ocean Model) is a finite volume ocean model suitable for simulating ocean flows in complex terrain. It can consider the influence of tides, wind, and other factors on ocean flows. SWAVE (Nearshore Wave Model) is a wave model used to simulate waves from the deep sea to nearshore areas, capable of simulating wave generation, propagation, and attenuation. The WRF-FVCOM-SWAVE coupled model achieves coordinated simulation of wind, waves, and currents through data interaction between modules. The WRF model provides atmospheric boundary conditions, such as wind speed, temperature, and air pressure, which are then passed to the FVCOM and SWAVE models. Sea surface temperature and ocean flow information calculated by the FVCOM model are fed back to the WRF to influence atmospheric simulation. Wave elements calculated by the SWAVE model are transferred to the FVCOM model, which uses them to calculate ocean currents under the influence of radiation stress and sea surface stress. Subsequently, the FVCOM model calculates ocean currents and water surface elevation results, which are then transferred to the SWAVE model for wave calculations in the next time step. WRF, FVCOM, and SWAVE achieve bidirectional nested dynamic coupling through shared boundary and state variables (wind field, current field, wave stress), enabling efficient simulation of the interaction processes between wind, waves, and currents and providing reliable marine meteorological environmental forecast information.

[0029] S12: Construct a deep learning correction model; acquire historical actual marine meteorological environment observation data and historical marine meteorological environment forecast information;

[0030] A deep learning correction model is trained using historical marine meteorological environment forecast information and historical actual marine meteorological environment observation data.

[0031] The marine meteorological and environmental forecast information is corrected by using a trained deep learning correction model.

[0032] As one example, a deep learning correction model optimizes forecast results through a deep learning model, further improving model accuracy. This process utilizes existing observational data and the errors generated by the numerical model to correct the model's forecast output, making the forecast results closer to actual observations. Using the error between historical numerical forecast data and observational data as the target variable, and wind-wave-current parameters as input, the input (wind-wave-current parameters) is combined with the target (model error) to train a deep learning model capable of predicting and correcting the errors of the wind-wave-current coupled numerical forecast model.

[0033] As one embodiment, the deep learning correction model is trained based on sample data; the sample data includes sample input data and sample output data, the sample input data includes sample flow feature parameters, sample wind feature parameters, and sample wave feature parameters, and the sample output data includes sample flow feature parameter errors, sample wind feature parameter errors, and sample wave feature parameter errors, and the sample errors are all obtained by subtracting the corresponding sample output data from the measured parameters of the samples respectively.

[0034] S13: Based on the prior information of distressed targets at sea and the corrected marine meteorological and environmental forecast information, construct and train a deep learning model of the initial distribution of distressed targets at sea based on the convolutional long short-term memory network algorithm; the prior information includes: the planned navigation route of distressed vessels at sea, the basic information of distressed vessels at sea, the basic information of personnel who have fallen into the water at sea, and feedback information from passing vessels.

[0035] Obtain real-time distress prior information; predict the real-time distress prior information using a deep learning model of the initial distribution of distressed targets at sea, and obtain the initial distribution set of targets;

[0036] As one example, prior information refers to the prior knowledge that predicts a distress event based on existing data and experience before the event occurs. It also includes the planned navigation route of a distressed vessel at sea, basic information of the distressed vessel at sea, basic information of persons who have fallen into the water at sea, and feedback information from passing vessels.

[0037] As one example, the deep learning model for the initial distribution of distressed targets at sea is a predictive model based on historical data and deep learning algorithms, used to predict the possible location distribution of distressed targets. This method trains the model by inputting a large amount of historical distress event data (including last reported location, last reported time, distress type, number of people in distress, etc.) and prior distress information. The model parameters are optimized by minimizing the error between the predicted location and the actual initial distribution location. Cross-validation is used to evaluate the model's generalization ability to ensure its performance on new data. The trained model can be used for actual predictions, providing the initial distribution under specific conditions to help search operations determine the potential initial distribution of targets. The output initial distribution is visualized as a probability density map, showing the potential risks in different areas, used for initial location during search operations. Wherein: the deep learning model for the initial distribution of distressed targets at sea is trained based on the initial distribution sample data of distressed targets at sea; the initial distribution sample data of distressed targets at sea includes distribution sample input data and distribution sample output data. The distribution sample input data consists of historical distress event data and prior information about the distress, including the last reported location, the last reported time, the type of distress incident, and the number of people in distress. The distribution sample output data includes the initial distribution location of distressed targets at sea.

[0038] S14: Based on wind-induced drift, current-induced drift, and wave-induced drift, construct a drift trajectory prediction model for distressed targets at sea, predict drift trajectories based on the initial distribution set of targets, and obtain the final distribution set;

[0039] As one example, a maritime distress target drift trajectory prediction model is used to predict the drift path of distressed targets in the marine environment. Based on dynamic factors such as ocean currents, wind speed, and waves, as well as the characteristics of the target itself, this model predicts the target's movement trajectory within a specific time period, thereby helping search personnel more accurately determine the target's current and future location and improving search efficiency. Using a Lagrange drift model, factors such as ocean currents, wind, and waves are transformed into the target's motion equations to derive its possible drift trajectory. Wind-induced drift considers the influence of wind on the target's drift; the action of wind causes changes in drift speed, which is modeled by considering wind speed and direction. Current-induced drift considers the influence of ocean current direction and speed on the drift path, obtaining ocean current data from a wind-wave-current coupling model to calculate the target's actual movement in the flowing environment. Wave-induced drift considers the periodic pressure exerted by waves on the drifting target, obtaining wave data from a wind-wave-current coupling model to calculate the impact of waves on the distressed target's location.

[0040] S15: Based on the final distribution set, combined with the convex hull algorithm, determine the search area.

[0041] As one example, the final distribution set refers to the set of information about the final position or state of distressed targets (such as floating objects, vessels, or personnel) in the ocean, generated after the initial distribution deep learning model and the drift trajectory prediction model of distressed targets at sea. The convex hull algorithm is used to calculate the convex hull of a set of points, which is the smallest convex polygon containing all predicted final distribution points.

[0042] As one example, the predicted distribution points are respectively , , .according to Sort the coordinates of the point set if If the coordinates are the same, then according to The coordinates are sorted. The sorted point set is traversed from left to right. For each point, it is checked whether the current point forms a right turn with the previous point and the second-to-last point (i.e., the three points are collinear or turn outwards). If a right turn is formed, the second-to-last point is removed; otherwise, the current point is added to the lower half of the convex hull, and the lower half of the convex hull is constructed progressively. After the lower half of the convex hull is constructed, the sorted point set is traversed in reverse, and the construction process of the lower half of the convex hull is repeated to construct the upper half of the convex hull. Finally, the lower and upper half of the convex hulls are merged to form a complete convex hull, which is used as the search area.

[0043] Step S2 includes:

[0044] S21: Obtain distress information for a target in distress at sea; distress information includes: last reported location, last reported time, type of distress, number of people in distress, and sea temperature;

[0045] S22: Based on distress information, combined with experience-based assessment and analytic hierarchy process, the urgency of the search operation is quantitatively assessed to obtain the urgency of the search operation.

[0046] S23: Obtain marine meteorological and environmental data, and calculate the marine risk index by combining empirical assessment methods and analytic hierarchy process; marine meteorological and environmental data include: sea surface wind speed, wave height, sea visibility, and current speed;

[0047] S24: Based on the K-means clustering method and combined with the marine risk index, the search area is divided into different sub-areas according to the risk level to obtain the marine risk level of the search area;

[0048] Risk levels include: extreme risk, high risk, medium risk, low risk, and no risk.

[0049] As one example, the comprehensive hierarchical analysis method is a multi-level analysis model that takes into account the combined influence of one or more factors of marine meteorological environment data such as sea surface wind speed, wave height, sea visibility, and current speed for different types of marine vessels. When calculating the sea area risk index, multiple hierarchical analysis models are used to calculate the risk index of the search sea area respectively, and the risk level of the search sea area is determined according to the principle of the highest risk level.

[0050] Among them, the K-means clustering method clusters each point in the searched sea area according to the risk index, determines the center of each cluster, and divides sea areas with similar risk levels into the same category, thereby realizing the graded assessment of sea area risk level.

[0051] Step S22 includes:

[0052] S221: Set the type of distress incident, the number of people in distress, and sea surface temperature as evaluation indicators for distress information, i.e. ;

[0053] Set the rating scale as ={Extremely urgent, Highly urgent, Moderately urgent, Mildly urgent};

[0054] Table 1. Classification of the urgency of the search operation

[0055]

[0056] S222: Construct an urgency assessment matrix using an empirical evaluation method;

[0057] Set evaluation indicators The degree of membership of each element in the evaluation set V , This indicates the evaluation indicators The k-th level gives the evaluation result as The proportion of people who were evaluated out of the total number of people evaluated, i.e., the evaluation indicator. The kth grade has the following comments: The degree;

[0058] Depend on Construct a single evaluation indicator set :

[0059]

[0060] in Indicates the first The parameters are divided into m levels. Characterization The probability of the kth urgency level; ; ;

[0061] S223: Use the analytic hierarchy process (AHP) to determine the weight set of the evaluation indicators. ;

[0062] urgency assessment indicators Construct the judgment matrix as follows:

[0063]

[0064] Among them, the judgment matrix has , , Properties; elements This indicates the influencing factor in relation to the urgency of the search operation. and The relative importance;

[0065] As one example, The values ​​are calculated using a 1-9 scale; specific scale values ​​and their meanings are shown in Table 2.

[0066] Table 2. Scale values ​​and meanings of matrix elements.

[0067]

[0068] Calculate the product of each row of the judgment matrix. , ;

[0069] calculate cube root , ;

[0070] vector Normalization to , ;

[0071] Calculate the largest eigenvalue. ;

[0072] Adopting consistency indicators The CI is compared with the average random consistency index RI to test whether the judgment matrix has satisfactory consistency, and the consistency test results are obtained.

[0073] As one example, the RI values ​​for matrices of orders 1 to 9 are shown in Table 3.

[0074] Table 3. Average random consistency index of matrices of order 1-9

[0075]

[0076] Based on the results of the consistency test, determine the weight set of the evaluation indicators. , representing the importance ranking of the three assessment indicators—sea temperature, type of distress, and number of people in distress—to the urgency of the search operation;

[0077] S224: A comprehensive evaluation is conducted based on the weight set of the evaluation indicators to obtain the probability of the search urgency.

[0078] Search urgency probability matrix ,in When planning this search operation, The probability of occurrence at each level of urgency; This represents a set of evaluation indicators for a single assessment.

[0079] S225: Calculate the urgency of the search operation based on the probability of search urgency;

[0080] Set a risk level ratio threshold L. If the ratio of the largest element to the second largest element in the search urgency probability matrix A is greater than the threshold L, then the urgency level corresponding to the largest element in the search urgency probability matrix A is taken as the final urgency level of the search action.

[0081] If the ratio of the largest element to the second largest element in the urgency probability matrix A is less than the threshold L, then the final urgency of the search action is determined by the median of each urgency level in A.

[0082] Step S23 includes:

[0083] S231: Set marine meteorological and environmental data as an evaluation indicator, i.e. ;

[0084] Set the risk assessment level set as ;

[0085] Table 4. Classification of Search Force Navigation Search Risk Assessment Levels

[0086]

[0087] S232: Repeat steps S222 to S225 to obtain the marine risk index, as follows:

[0088] An urgency assessment matrix was constructed using an empirical evaluation method.

[0089] The weight set of evaluation indicators was determined by using the analytic hierarchy process (AHP) combined with an urgency assessment matrix.

[0090] A comprehensive evaluation is conducted based on the weight set of the assessment indicators to obtain the probability of the search risk level for all grid points within the search area.

[0091] By searching for the probability of risk levels, the risk index of all grid points in the search area is calculated, which is the sea area risk index.

[0092] Step S24 includes:

[0093] S241: Determine the number of clusters required by using the elbow rule and combining it with the risk index of the search area. ;

[0094] As one example, the number of clusters It depends on the size of the search area. A reasonable approach... The value range is usually 1 to 10.

[0095] For each Value, Execution Mean clustering, calculating the risk index for each grid point. Its corresponding cluster center risk index Sum of squared differences As a cost function for clustering;

[0096] The sum of squares of the calculated differences With the corresponding The values ​​are plotted as a curve, with the horizontal axis being... Values, with the vertical axis as value;

[0097] As one example, in the drawn diagram, the search for " The point where the rate of decrease in value slows significantly is typically shaped like an "elbow." The value represents the optimal number of clusters.

[0098] The K value at which the curve in the graph appears is selected as the number of clusters.

[0099] S242: Using random sampling, select One data point is used as the initial cluster center;

[0100] As one implementation example, random sampling ensures that each individual has an equal probability of being selected, and the selection of each individual in the sample is independent of each other and unaffected by other individuals. A random number generator is used to ensure the randomness of the sample selection process.

[0101] S243: Based on each data point Risk index ( ) and the risk index of each cluster center ( The difference Each data point is assigned to the nearest cluster center;

[0102] S244: For each cluster Calculate all points Average risk index Update the cluster center values;

[0103]

[0104]

[0105] S245: Repeat steps S243 and S244 until the cluster center value is reached. The iteration terminates if the condition remains unchanged.

[0106] Based on clustering, the search area is divided into sub-areas with different risk levels;

[0107] S246: In accordance with the principle of the highest risk level, the highest risk level appearing in each sub-sea area shall be taken as the sea area risk level of the search area.

[0108] As one implementation example, based on marine meteorological environmental data such as sea surface wind speed, wave height, sea visibility, and current speed, a comprehensive analytic hierarchy process (AHP) is used to calculate a marine risk index. Then, the search area is divided into different sub-areas according to risk level using K-means clustering. The AHP is a multi-level analysis model that considers the combined influence of one or more factors of marine meteorological environmental data such as sea surface wind speed, wave height, sea visibility, and current speed for different types of vessels. When calculating the marine risk index, multiple AHP models are used to calculate the risk index of the search area separately, and the risk level of the search area is determined according to the highest risk level. The K-means clustering method clusters points in the search area according to the risk index, determines the center of each cluster, and groups areas with similar risk levels into the same category, thereby achieving a graded assessment of marine risk levels.

[0109] Step S3 includes:

[0110] S31: Determine the search resource;

[0111] S311: Obtain the risk resistance levels of all available search aircraft and search vessels around the search area;

[0112] S312: Screening search aircraft and search vessels based on sea area risk level and risk resistance level;

[0113] S313: Calculate the round-trip time of the search aircraft between the aircraft base and the search area, as follows:

[0114]

[0115] in To search for the round-trip time of the aircraft, The distance between the flight base and the search area. The maximum speed of the search aircraft; The impact coefficient of the risk level. This indicates the risk level of a gale warning during the flight's round trip.

[0116] Obtain the maximum flight time of the search aircraft;

[0117] Remove search aircraft whose maximum endurance is less than the round-trip time, perform a second screening of search aircraft, and retain the final usable search aircraft;

[0118] S314: The search aircraft and search vessels that are ultimately retained are the final search resources;

[0119] S32: Establish a resource search decision model; through the resource search decision model, combined with the finally determined search resources, obtain a comprehensive index of the decision scheme;

[0120] The resource search decision model is a multi-objective integer nonlinear programming model that aims to minimize search action time, minimize search resource cost, and maximize search action security. It is expressed as follows:

[0121]

[0122] in This represents the comprehensive index of the decision-making options. For search time, To reduce search costs, For search security, , , They are respectively with , , The corresponding weights are calculated based on the urgency of the search operation;

[0123] , ,as well as Determined by the following function:

[0124]

[0125] in, This indicates the total number of available ships after the initial screening. This indicates the total number of available aircraft after initial screening. Indicates the first Search vessels, , Indicates the first A search aircraft, ; Indicates the first The search status of the search vessels. , Time indicates the first Several search vessels participated in the search operation. Time indicates the first The search vessels did not participate in the search operation. Indicates the first The search status of the search aircraft. , Time indicates the first Search aircraft participated in the search operation. Time indicates the first The search aircraft will not participate in the search operation. Indicates the first The time required for the ships to travel at full speed to the search area. Indicates the first The search capabilities of the ships Indicates the first The search capabilities of the aircraft Indicates the area of ​​the search area. Indicates the first The time required for an aircraft to travel between the flight base and the search area. Indicates the first The maximum flight time of an aircraft. Indicates the first The number of times search aircraft were deployed throughout the entire search operation. Indicates the first The cost of dispatching a ship, Indicates the first The cost of dispatching an aircraft. Indicates the first Fuel consumption rate of a ship Indicates the first The fuel consumption rate of an aircraft Indicates the first The unit price of fuel used by each ship. Indicates the first The unit price of fuel used by the aircraft. Indicates the first The risk resistance level of a ship For the first The risk resistance level of an aircraft Based on the risk level of the search area, No. The distance between the aircraft and the area to be searched. Indicates the risk level of the flight's round trip. Indicates the first The maximum speed of the aircraft The impact coefficient of the risk level. This represents the maximum number of search vessels that the search area can accommodate. This is the maximum number of aircraft that can be searched.

[0126] , , The allocation method is determined by the urgency of the search operation and is as follows:

[0127] When the search operation is classified as extremely urgent , , The values ​​are 0.9, 0.05, and 0.05 respectively.

[0128] When the search operation is classified as highly urgent , , They are 0.8, 0.1, and 0.1 respectively;

[0129] When the search operation is classified as moderately urgent , , The values ​​are 0.7, 0.15, and 0.15 respectively.

[0130] When the search operation is classified as mildly urgent, , , The values ​​are 0.6, 0.2, and 0.2 respectively.

[0131] S33: The search resource planning scheme with the smallest comprehensive index of the decision scheme output by the search resource decision model is determined as the optimal search resource planning scheme.

[0132] As one example, the search resource planning scheme derived from the search resource decision model specifically includes the number of search aircraft and the number of search ships. Based on the search resource decision model, a comprehensive index of the decision scheme for the search resource planning scheme is calculated, and the search resource planning scheme with the smallest comprehensive index is determined as the optimal search resource planning scheme.

[0133] Step S4 includes:

[0134] S41: Based on the optimal search resource planning scheme and the divided sub-regions, determine the basic information of the search force; the basic information of the search force includes: the risk level of the search sub-region, the area of ​​the search sub-region, the risk resistance level of the search force, and the maximum workload of the search force.

[0135] As one example, the area of ​​sub-regions with different sea area risk levels is calculated; based on the optimal search resource planning scheme, the maximum workload that the search force can complete within the search duration is calculated, specifically expressed as the search force's search capability multiplied by the working duration.

[0136] S42: The work layout of search forces is allocated based on a greedy algorithm, including: the greedy algorithm allocates search forces in descending order of risk level according to the matching priority between the risk resistance level of search forces and the risk level of search sub-regions;

[0137] S421: Let the set of input data sub-regions be... ,in Indicates the first Risk level of each search sub-region Indicates the first Area of ​​each search sub-region; input data search force set ,in Indicates the first Each search force's risk resistance level, Indicates the first The maximum workload of a search force;

[0138] S422: Search by sub-region risk level sorted in descending order According to the risk level of search capabilities sorted in descending order ;

[0139] S423: Initialize the search resource allocation result set traverse and search for the collection of forces Regarding the current search capabilities Find the first one subregion ;

[0140] like Then the area is allocated Give the sub-region the update. ;

[0141] like Then the area is allocated Give the search power and update it. ;

[0142] After a thorough search, the forces were gathered. Output a set of search resource allocation results. That is, the assigned area and corresponding area of ​​each search force, i.e., the working sea area of ​​the search force;

[0143] S43: Based on the search resource allocation result set The optimal global path planning for the search forces is obtained by performing optimal global path planning for the search forces.

[0144] As one example, the global search methods mainly include sector search, extended square search, and parallel line search, and are selected according to the following rules:

[0145] The fan-shaped search method is mainly suitable for situations where the target location is accurate or the search area is small. For search vessels, the fan-shaped search radius is typically 2 to 5 nautical miles. If the target is not found after completing one fan-shaped search, the fan is rotated and lowered by half the radius of the first search to conduct a second fan-shaped search.

[0146] The extended square search is suitable when the target location is in a relatively close area. In this method, the search starting point is always the reference position, and the search expands outwards in concentric squares, thus uniformly covering the area centered on the reference point. However, if the reference point is not a point but a short line, it should be changed to an outward-expanding rectangle. The length of the two search lines is equal to the distance between them, and every two subsequent search segments are increased by one search line distance. If continuous searching is being conducted within the same area, the search line is generally turned 45° to continue the search.

[0147] Parallel line search is suitable for situations where the location of the target is highly uncertain and uniform coverage of a wide area is required. First, a search area is set, and the spacing of the search lines is determined according to the site conditions. The search equipment uses one corner of the search area as the starting point for the search. The starting point is usually located within the search rectangle at a distance of 1 / 2 the distance between the two right-angled sides of the search line. Then, the search is carried out back and forth along the long side of the rectangle while maintaining the spacing.

[0148] This application also discloses an electronic device. (See reference...) Figure 2 , Figure 2 This is a schematic diagram of the structure of an electronic device disclosed in an embodiment of this application. The electronic device 500 may include: at least one processor 501, at least one network interface 504, a user interface 503, a memory 505, and at least one communication bus 502.

[0149] The communication bus 502 is used to enable communication between these components.

[0150] The user interface 503 may include a display screen, and optionally, the user interface 503 may also include a standard wired interface or a wireless interface.

[0151] The network interface 504 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface).

[0152] This application also discloses a computer-readable storage medium storing multiple instructions adapted for loading by a processor to execute the above-described method, apparatus, and medium for assisting decision-making in searching for distressed targets at sea.

[0153] The above are merely exemplary embodiments of this disclosure and should not be construed as limiting the scope of this disclosure. Any equivalent changes and modifications made in accordance with the teachings of this disclosure shall still fall within the scope of this disclosure.

[0154] This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not described in this disclosure. The specification and embodiments are to be considered exemplary only, and the scope and spirit of this disclosure are defined by the claims.

Claims

1. A method for assisting decision-making in searching for distressed targets at sea, characterized in that, The method includes the following steps: S1: Obtain marine meteorological and environmental forecasts for the distress area and surrounding waters to determine the search area; S2: Calculate the risk level of the search area and the urgency of the search operation; S3: Obtain search resources; Establish a search resource decision model, and combine the urgency of the search operation, the risk level of the sea area, and the search resources to obtain the best search resource planning scheme; Step S3 includes: S31: Determine the search resource; S311: Obtain the risk resistance levels of all available search aircraft and search vessels around the search area; S312: Screening search aircraft and search vessels based on sea area risk level and risk resistance level; S313: Calculate the round-trip time of the search aircraft between the aircraft base and the search area; Obtain the maximum flight time of the search aircraft; Remove search aircraft whose maximum endurance is less than the round-trip time, perform a second screening of search aircraft, and retain the final usable search aircraft; S314: The search aircraft and search vessels that are ultimately retained are the final search resources; S32: Establish a resource search decision model; through the resource search decision model, combined with the finally determined search resources, obtain a comprehensive index of the decision scheme; The resource search decision model is a multi-objective integer nonlinear programming model that aims to minimize search action time, minimize search resource cost, and maximize search action security. It is expressed as follows: Where G represents the comprehensive index of the decision scheme, T is the search time, C is the search cost, S is the search security, and μ1, μ2, and μ3 are the weights corresponding to T, C, and S, respectively, calculated from the urgency of the search action; S33: The search resource planning scheme with the smallest comprehensive index of the decision scheme output by the search resource decision model is determined as the best search resource planning scheme; S4: Based on the risk level of the sea area, the search area is divided into sub-regions. Combined with the best search resource planning scheme, the working sea area of ​​the search force is obtained and the optimal global path planning of the search force is determined.

2. The method for assisting decision-making in searching for distressed targets at sea as described in claim 1, characterized in that, Step S1 includes: S11: Establish a wind-wave-current coupled numerical forecasting model and obtain marine meteorological environment forecast information for the distress area and nearby sea areas through the wind-wave-current coupled numerical forecasting model; The wind-wave-current coupled numerical prediction model includes: an atmospheric module, a wave module, and an ocean current module; S12: Construct a deep learning correction model; acquire historical actual marine meteorological environment observation data and historical marine meteorological environment forecast information; A deep learning correction model is trained using historical marine meteorological environment forecast information and historical actual marine meteorological environment observation data. The marine meteorological and environmental forecast information is corrected by using a trained deep learning correction model. S13: Based on the prior information of distressed targets at sea and the corrected marine meteorological and environmental forecast information, construct and train a deep learning model of the initial distribution of distressed targets at sea based on the convolutional long short-term memory network algorithm; the prior information includes: the planned navigation route of distressed vessels at sea, the basic information of distressed vessels at sea, the basic information of personnel who have fallen into the water at sea, and feedback information from passing vessels. Obtain real-time distress prior information; predict the real-time distress prior information using a deep learning model of the initial distribution of distressed targets at sea, and obtain the initial distribution set of targets; S14: Based on wind-induced drift, current-induced drift, and wave-induced drift, construct a drift trajectory prediction model for distressed targets at sea, predict drift trajectories based on the initial distribution set of targets, and obtain the final distribution set; S15: Based on the final distribution set, combined with the convex hull algorithm, determine the search area.

3. The method for assisting decision-making in searching for distressed targets at sea as described in claim 1, characterized in that, Step S2 includes: S21: Obtain distress information for a target in distress at sea; distress information includes: last reported location, last reported time, type of distress, number of people in distress, and sea temperature; S22: Based on distress information, combined with experience-based assessment and analytic hierarchy process, the urgency of the search operation is quantitatively assessed to obtain the urgency of the search operation. S23: Obtain marine meteorological and environmental data, and calculate the marine risk index by combining empirical assessment methods and analytic hierarchy process; marine meteorological and environmental data include: sea surface wind speed, wave height, sea visibility, and current speed; S24: Based on the K-means clustering method and combined with the marine risk index, the search area is divided into different sub-areas according to the risk level to obtain the marine risk level of the search area; Risk levels include: extreme risk, high risk, medium risk, low risk, and no risk.

4. The method for assisting decision-making in searching for distressed targets at sea as described in claim 3, characterized in that, Step S22 includes: S221: Set the distress incident type, number of people in distress, and sea temperature as evaluation indicators, i.e., U={u1,u2,u3}={sea temperature, distress incident type, number of people in distress}; Set the evaluation level set as V = {v1, v2, v3, v4} = {extremely urgent, highly urgent, moderately urgent, mildly urgent}; S222: Construct an urgency assessment matrix using an empirical evaluation method; Let the evaluation index u i The degree of membership r of each element in the evaluation set V ik,j r ik,j This indicates the evaluation index u i The k-th level gives an evaluation result of v j The proportion of people who meet the criteria out of the total number of people evaluated, i.e., the evaluation indicator u. i The kth grade has a comment of v j The degree; By r ik,j The evaluation set R constitutes a single evaluation indicator i : Where m represents that the i-th parameter is divided into m levels, r ik,j Characterization u i The probability of the kth urgency level; i = 1, 2, 3; j = 1, 2, 3, 4; S223: Use the analytic hierarchy process (AHP) to determine the weight set W = [W1; W2; W3] for the evaluation indicators; For the urgency assessment index U, a judgment matrix is ​​constructed as follows: Wherein, the judgment matrix has C ij >0、 C ii =1 property; element C ij The impact factor u represents the degree of urgency of the search operation. i and u j The relative importance; Calculate the product M of each row of the judgment matrix. i , Calculate M i cube root vector Normalized to W = [W1; W2; W3], Calculate the largest eigenvalue. Adopting consistency indicators The CI is compared with the average random consistency index RI to test whether the judgment matrix has satisfactory consistency, and the consistency test results are obtained. Based on the results of the consistency test, the weight set W = [W1; W2; W3] of the evaluation indicators is determined, representing the importance ranking weights of the three evaluation indicators of sea surface temperature, distress type, and number of people in distress to the urgency of the search operation. S224: A comprehensive evaluation is conducted based on the weight set of the evaluation indicators to obtain the probability of the search urgency. The search urgency probability matrix A = R × W = [a1; a2; a3; a4], where a j V represents the probability of each urgency level occurring when planning this search operation; R represents the evaluation set of a single assessment index. S225: Calculate the urgency of the search operation based on the probability of search urgency; Set a risk level ratio threshold L. If the ratio of the largest element to the second largest element in the search urgency probability matrix A is greater than the threshold L, then the urgency level corresponding to the largest element in the search urgency probability matrix A is taken as the final urgency level of the search action. If the ratio of the largest element to the second largest element in the urgency probability matrix A is less than the threshold L, then the final urgency of the search action is determined by the median of each urgency level in A.

5. The method for assisting decision-making in searching for distressed targets at sea as described in claim 4, characterized in that, Step S23 includes: S231: Set marine meteorological environmental data as evaluation indicators, i.e., U={u1,u2,u3,u4}={sea surface wind speed, wave height, sea visibility, current speed}; Let the risk assessment level set be V = {v1, v2, v3, v4, v5} = {Extremely high risk, high risk, medium risk, low risk, no risk}; S232: Repeat steps S222 to S225 to obtain the marine risk index, as follows: An urgency assessment matrix was constructed using an empirical evaluation method. The weight set of evaluation indicators was determined by using the analytic hierarchy process (AHP) combined with an urgency assessment matrix. A comprehensive evaluation is conducted based on the weight set of the assessment indicators to obtain the probability of the search risk level for all grid points within the search area. By searching for the probability of risk levels, the risk index of all grid points in the search area is calculated, which is the sea area risk index.

6. The method for assisting decision-making in searching for distressed targets at sea as described in claim 3, characterized in that, Step S24 includes: S241: Determine the number of clusters K to be clustered by using the elbow rule and combining it with the risk index of the search area; For each K value, perform K-means clustering to calculate the risk index for each grid point. Its corresponding cluster center risk index Sum of squared differences As a cost function for clustering; Plot the calculated sum of squared differences H against the corresponding K values ​​on a curve, with the horizontal axis representing the K values ​​and the vertical axis representing the H values. The K value at which the curve in the graph appears is selected as the number of clusters. S242: Use random sampling to select K data points as the initial cluster centers; S243: Based on each data point P i Risk Index Risk index of each cluster center The difference Assign each data point to the nearest cluster center; S244: For each cluster AR i Calculate x for all points i The average risk index AS i Update the cluster center values; x i =x1,x2,x3…x n S245: Repeat steps S243 and S244 until the cluster center value AS is reached. i The iteration terminates if the condition remains unchanged. Based on clustering, the search area is divided into sub-areas with different risk levels; S246: In accordance with the principle of the highest risk level, the highest risk level appearing in each sub-sea area shall be taken as the sea area risk level of the search area.

7. The method for assisting decision-making in searching for distressed targets at sea as described in claim 1, characterized in that, Step S4 includes: S41: Based on the optimal search resource planning scheme and the divided sub-regions, determine the basic information of the search force; the basic information of the search force includes: the risk level of the search sub-region, the area of ​​the search sub-region, the risk resistance level of the search force, and the maximum workload of the search force. S42: The work layout of search forces is allocated based on a greedy algorithm, including: the greedy algorithm allocates search forces in descending order of risk level according to the matching priority between the risk resistance level of search forces and the risk level of search sub-regions; S421: Let the set of input data sub-regions be Y = {(r i ,s i )}, where r i s represents the risk level of the i-th search sub-region. i Represents the area of ​​the i-th search sub-region; input data: search force set F = {(r j ,a j )}, where r j Indicates the risk resistance level of the j-th search force, a j This represents the maximum workload of the j-th search force; S422: Search by sub-region risk level r i Sort by descending order Y; sort by search strength risk level r j Arrange F in descending order; S423: Initialize the search resource allocation result set R1, traverse the search force set F, and for the current search force (r j ,a j Find the first r i ≤r j subregion (r) i ,s i ); If a j ≤s i Then the area a is allocated j Give the sub-region a value and update s. i =s i -a j ; If a j >s i Then the allocated area s i Give the search power and update a j =a j -s i ; After traversing the search force set F, the search resource allocation result set R1 is output, which is the allocation area and corresponding area of ​​each search force, i.e. the working sea area of ​​the search force. S43: Based on the search resource allocation result set R1, perform optimal global path planning for the search forces to obtain the optimal global path planning for the search forces.

8. An electronic device, characterized in that, The device includes a processor, a memory, a user interface, and a network interface. The memory is used to store instructions, the user interface and the network interface are used to communicate with other devices, and the processor is used to execute the instructions stored in the memory to cause the electronic device to perform the method as described in any one of claims 1-7.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores instructions that, when executed by a computer, perform the method as described in any one of claims 1-7.

Images