Intelligent ship trajectory analysis system and method thereof

Abstract

The invention relates to the technical field of intelligent shipping, and discloses a ship intelligent track analysis system and method, and the system comprises a data collection module, a track modeling module, a track anomaly detection module, a track optimization module, a global track reconstruction module and a feedback and decision support module. The data acquisition module is used for acquiring operation basic data, environment information and multi-source fusion data of a ship. By adopting the technical scheme of multi-module cooperation, comprehensive monitoring and intelligent analysis of the ship moving trajectory are realized, and the problem of weak data integration capability in the prior art is solved; by introducing the recurrent neural network and the long and short term memory network, the accuracy of trajectory modeling and prediction is improved; a global track reconstruction module is used for generating a multi-ship cooperative optimal path, and the defect of dispersed route planning is overcome; meanwhile, a visual and flexible feedback and decision support module is provided, the path adjustment response speed is optimized, and the practicability and the intelligent level of the system are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent shipping, and specifically to a ship intelligent trajectory analysis system and method thereof. Background Art

[0002] Existing ship trajectory monitoring systems mainly rely on a single data source, such as GPS or AIS (Automatic Identification System for Ships), and cannot comprehensively integrate ship operation status and environmental data (such as wind speed, wave height, and flow velocity). This single data mode is difficult to cope with complex and changeable sea conditions, limits the real-time response ability of trajectory analysis to the external environment, and results in the inability to accurately identify potential risks or abnormal behaviors during actual navigation. At the same time, most existing trajectory modeling and prediction methods are based on rule models or simple statistical methods, and cannot handle complex time series and non-linear trajectory features. In the scenario of multi-ship collaborative operation, this method cannot provide sufficient modeling accuracy, often resulting in lagging trajectory prediction or failure of anomaly detection.

[0003] In terms of path optimization, existing technologies generally perform independent optimization with a single ship as the unit, lacking overall consideration of multi-ship collaboration. In navigation-intensive areas or multi-ship co-navigation scenarios, this independent path planning is prone to path conflicts and navigation risks, significantly reducing the overall operation efficiency. In addition, the path optimization results of existing systems are usually presented in a static form, and the operations of users are limited by the fixed optimization results, lacking flexible interaction means. The system cannot quickly adjust the path planning in dynamic sea conditions, with slow feedback speed, and is difficult to meet the high-efficiency operation requirements of ships in complex environments.

[0004] These problems lead to the difficulty of existing technologies in meeting the actual shipping needs in terms of data fusion ability, trajectory modeling accuracy, multi-ship collaborative optimization, and user interaction experience, seriously restricting their applicability and safety in complex shipping scenarios. Summary of the Invention

[0005] Aiming at the deficiencies of the existing technology, the present invention provides a ship intelligent trajectory analysis system and method thereof, which solve the problems of insufficient accuracy of ship trajectory modeling, weak data integration and anomaly detection capabilities, lack of multi-ship collaborative optimization, and inflexible path planning feedback mechanism in the existing technology.

[0006] To achieve the above objectives, the present invention is realized through the following technical solutions: A ship intelligent trajectory analysis system, including a data acquisition module for collecting basic operation data, environmental information, and multi-source fusion data of the ship; A trajectory modeling module for establishing a ship operation trajectory model based on the multi-dimensional data obtained by the data acquisition module; A trajectory anomaly detection module for identifying yaw, collision risks, and abnormal behaviors based on the trajectory model generated by the trajectory modeling module; A trajectory optimization module, configured to generate path adjustment suggestions according to the detection results of the trajectory anomaly detection module; A global track reconstruction module, configured to reconstruct and optimize the overall track of the ship group in combination with cluster collaborative analysis; A feedback and decision support module, configured to feedback the trajectory analysis results and optimization suggestions to the ship intelligent management system to support route adjustment decisions.

[0007] Preferably, the data acquisition module includes: A ship operation state acquisition unit, configured to acquire the speed, heading and position data of the ship; An environmental data acquisition unit, configured to obtain weather information, sea condition data and obstacle information; A data fusion unit, configured to integrate multi-source data and output a unified data format.

[0008] Preferably, the trajectory modeling module predicts the ship trajectory based on a time series algorithm, and the prediction process includes: Performing time series analysis on the ship operation state and environmental data by using a recurrent neural network; Predicting the next position point of the ship according to the historical trajectory and the current state.

[0009] Preferably, the trajectory anomaly detection module includes: A yaw detection unit, configured to calculate the deviation degree between the current position of the ship and the target route; A collision risk assessment unit, configured to calculate a collision risk value based on the distance and relative speed between the current position of the ship and the obstacle; An abnormal behavior detection unit, configured to identify abnormal trajectory patterns through a deep learning model.

[0010] Preferably, the trajectory anomaly detection module includes: A yaw detection unit, configured to calculate the deviation degree between the current position of the ship and the target route; A collision risk assessment unit, configured to calculate a collision risk value based on the distance and relative speed between the current position of the ship and the obstacle; An abnormal behavior detection unit, configured to identify abnormal trajectory patterns through a deep learning model.

[0011] A ship intelligent trajectory analysis method, including the following steps: S1: Acquiring the basic ship operation data, environmental information and multi-source fusion data through the data acquisition module; S2: Establishing a ship operation trajectory model through the trajectory modeling module; S3: Detecting the yaw and collision risks in the trajectory through the trajectory anomaly detection module; S4: Generate path adjustment suggestions through the trajectory optimization module; S5: Based on multi-ship data, conduct collaborative analysis through the global track reconstruction module and reconstruct the optimized route; S6: Feed back the trajectory optimization results to the ship intelligent management system through the feedback and decision support module for display and decision adjustment.

[0012] Preferably, the trajectory modeling in step S2 includes: Establish a trajectory prediction model based on the time series data of ship operation; Update the trajectory model in real time through the long short-term memory network; Output the predicted trajectory points and status information of the ship.

[0013] Preferably, the path optimization in step S4 includes the following steps: Combine the detection results of the trajectory anomaly detection module to calculate the risk weight; Use the dynamic programming algorithm to generate the shortest path and the lowest risk path; Output the optimized path for the reference of the ship intelligent management system.

[0014] Preferably, the global track reconstruction in step S5 includes the following steps: Establish a global collaboration model based on multi-ship data; Calculate the optimal distribution of all ships in the target area; Dynamically adjust the routes of each ship to maximize the overall operation efficiency.

[0015] Preferably, the feedback and decision support in step S6 include: Display the trajectory optimization results and risk warning information through a graphical interface; Provide an operation instruction interface for users to adjust the operation strategy of the ship.

[0016] The present invention provides a ship intelligent trajectory analysis system and method. It has the following beneficial effects: 1. The present invention adopts a technical solution of multi-module collaboration. Through functional modules such as data collection, trajectory modeling, anomaly detection, optimization and reconstruction, it realizes the comprehensive monitoring and intelligent analysis of the ship operation trajectory. Compared with the technical solutions that rely on a single module or independent functional modules in the prior art, it solves the deficiencies of the traditional system such as weak data integration ability, single trajectory analysis and lack of real-time dynamic adjustment.

[0017] 2. The present invention realizes accurate modeling of ship trajectories and prediction of future trajectories by introducing deep learning models of recurrent neural networks (RNNs) and long short-term memory networks (LSTMs), effectively improving the accuracy and real-time performance of trajectory analysis. Compared with the prediction methods based on simple rule models in the prior art, it solves the problems of insufficient processing ability for complex sea condition data and low trajectory prediction accuracy.

[0018] 3. The present invention adopts a global track reconstruction module, combines multi-objective optimization and dynamic programming algorithms to generate an optimal path distribution for multi-ship collaboration, significantly improving the overall operation efficiency and safety of the ship group. Compared with the technical solutions of single-ship optimization or lack of multi-ship collaboration in the prior art, it solves the problems of frequent path conflicts, scattered route planning, and lack of intelligence.

[0019] 4. The present invention provides an intuitive and flexible feedback and decision support module. Combining a graphical interface and various interaction methods, it can dynamically adjust path planning and optimize the operation plan in real time. Compared with the systems in the prior art with a single feedback process and limited decision support functions, it solves the deficiencies of poor user operation experience and slow response speed for path adjustment, and at the same time significantly improves the practicality and intelligence level of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 It is a schematic diagram of the framework process of the system in the present invention; Figure 2 It is a schematic diagram of the step process of the invention in the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0021] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the specification of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0022] Please refer to the attached Figure 1 , the embodiments of the present invention provide a ship intelligent trajectory analysis system, aiming to solve the problems of real-time performance, accuracy, and collaboration in ship trajectory monitoring and optimization. Through the full-process intelligent design of data collection, trajectory modeling, anomaly detection, optimization, and feedback, it realizes the efficient and safe operation of single ships and multi-ship groups.

[0023] The system includes: Data Acquisition Module In the ship intelligent trajectory analysis system of the present invention, the data acquisition module is the basic link for the operation of the entire system, mainly used to obtain multi-source data related to ship operation in real time. The data acquisition module not only undertakes the acquisition of basic data, but also needs to perform fusion processing on the multi-dimensional data collected to meet the input requirements of the subsequent trajectory modeling module. Generally, the data acquisition module needs to cooperate closely with on-board equipment, external environment perception equipment, and remote communication modules to achieve dynamic acquisition and processing of multi-level data. In the context, the output data of this module is directly supplied to the trajectory modeling module and becomes the input basis for trajectory prediction and subsequent analysis.

[0024] In this embodiment, the data acquisition module mainly consists of the following units: a ship operation state acquisition unit, an environmental data acquisition unit, and a data fusion unit.

[0025] Specifically, the ship operation state acquisition unit is used to obtain the dynamic operation data of the ship, including but not limited to information such as speed, position, heading, and acceleration. These data can be collected in real time through hardware such as on-board navigation equipment (such as GPS modules), inertial navigation systems (INS), and acceleration sensors. For example, the position data can be obtained in the form of geographical coordinates through the on-board GPS module, expressed as (x, y, z), where x and y are the plane coordinates on the earth's surface and z is the altitude; the speed data can be calculated by the INS device and expressed as a vector where v x , v y , v z are the components along the x, y, and z directions respectively.

[0026] The speed data in the operation state acquisition unit can also be verified by a Doppler radar to improve data accuracy. Specifically, the Doppler radar calculates the relative speed v of the ship relative to the target point by measuring the frequency offset Δf r , and the formula is: where c is the speed of light and f 0 is the frequency of the radar emission wave. After the above data is synchronized, it is used as the compensation value of the speed.

[0027] In some embodiments, the acceleration data of the operation state acquisition unit can also be collected by an acceleration sensor array installed at different positions on the hull and filtered based on a multi-point sampling algorithm to reduce the interference of environmental noise.

[0028] The environmental data acquisition unit is used to sense the external environmental information around the ship, including weather, sea conditions, and surrounding obstacle data. Specifically, weather data can be obtained through a meteorological satellite receiving module, mainly including wind speed, wind direction, air pressure, and precipitation; sea condition data is collected by on-board wave monitoring equipment, mainly including wave height, wave period, and flow velocity. The relevant parameters can be expressed as follows: Wave height H, unit: meter; Wave period T, unit: second; Flow velocity U = (u x , u y ), where u x and u y are the horizontal and vertical components respectively.

[0029] The obstacle data in the environmental data acquisition unit can be jointly collected by an automatic identification system (AIS) and on-board radar. AIS is used to identify the basic information of surrounding ships (such as position and speed), and on-board radar is used to detect non-identifiable obstacles (such as floating objects or underwater terrain protrusions). In one implementation, the distance d detected by the radar can be calculated by the formula: where Δt is the time difference between signal transmission and reception.

[0030] The data fusion unit is an important part of the data acquisition module. Its main function is to perform spatio-temporal alignment and format unification processing on the operation state data and environmental data to meet the input requirements of the trajectory modeling module. Generally, the data fusion unit ensures the validity of various data within the same time window through timestamp alignment technology. In addition, to improve the robustness of data fusion, the data fusion unit can introduce a fusion algorithm based on Kalman Filter. The specific steps include: Initial state estimation: Assume the initial state is where x 0 is the initial position and v 0 is the initial velocity; State prediction: Predict the state at the next moment according to the motion model; Measurement update: Combine the collected data with the predicted state and correct the estimated value through the weight update formula: where K k is the Kalman gain, z k is the measurement value, and H is the measurement matrix.

[0031] The data fusion unit can also integrate machine learning algorithms, use historical data to train a regression model, and predict and compensate for missing or abnormal data. For example, through the random forest regression algorithm, the short-term missing course data can be completed based on the collected speed and position data.

[0032] After being processed, the output data of the data acquisition module is unified into an input format that the trajectory modeling module can directly process, specifically a multi-dimensional time series matrix: Among them, each row represents the multi-dimensional data at a time point.

[0033] Through the detailed design and implementation of the above data acquisition module, comprehensive and accurate input data can be provided for trajectory modeling and subsequent modules, ensuring the real-time performance and efficiency of the overall system operation The trajectory modeling module is a key component connecting the preceding and the following in the present invention. Its main task is to establish an operation trajectory model of the ship based on the multi-source fusion data provided by the data acquisition module. Through this module, the current state of the ship can be described, and its future operation position can be predicted, providing data support for subsequent anomaly detection and path optimization. Generally, the trajectory modeling module needs to have the ability to process high-dimensional time series data while ensuring the real-time performance and accuracy of the modeling results. In the context logic, the input of this module is the standardized data matrix output by the data acquisition module, and the output is directly used by the trajectory anomaly detection module.

[0034] In this embodiment, the trajectory modeling module mainly uses a combination of machine learning and statistical methods for modeling. The module structure includes a data preprocessing unit, a time series modeling unit, and a trajectory prediction unit.

[0035] Specifically, the data preprocessing unit is used to clean, normalize, and time-align the input multi-source data. Generally, there may be certain delays or sampling frequency differences in the multi-source data in terms of time, so time synchronization processing is required. For example, in some embodiments, the low-frequency sampled data can be completed by interpolation methods. Suppose there is data missing at a certain time point t i The interpolation calculation formula can be expressed as: Among them, x(t i ) represents the data value corresponding to time t i .

[0036] As an option, to ensure the stability of model training, the data normalization operation is usually based on the min-max normalization method. The normalization formula is: where x is the original data, x min and x max are the minimum and maximum values of this feature respectively, and x′ is the data value after normalization.

[0037] The time series modeling unit is the core part of the trajectory modeling module and is used to generate the operation trajectory model of the ship based on the processed input data. In a possible implementation, this unit uses a recurrent neural network (RNN) for modeling. The RNN can process time series data and update the hidden state at each time step to capture the dependencies in the time dimension. Let the input data be a multi-dimensional vector sequence X = {x 1 , x 2 , …, x T}, where T is the number of time steps, and x t represents the input feature vector at the t-th moment. The update formula for the hidden state is: h t = f(W h h t-1 + W x x t + b h ) where h t is the hidden state at the t-th moment, W h and W x are weight matrices, b h is a bias vector, and f(·) is an activation function (such as tanh or ReLU).

[0038] To alleviate the problem of vanishing gradients in the processing of long sequences by the RNN, in some embodiments, a long short-term memory network (LSTM) is introduced for improvement. The LSTM adds memory cells in the hidden layer and realizes the memory and selective update of long-term dependence information through the combination of an input gate, a forget gate, and an output gate. Its calculation formula is: i t = σ(W i x t + U i h t-1 + b i ) f t = σ(W f x t + U f h t-1 + b f o t = σ(W o x t + U o h t-1 + b o ) ct = f t ⊙ c t-1 + i t ⊙ tanh(W c x t + U c h t-1 + b c ) h t = o t ⊙ tanh(c t ) where i t , f t and o t are the activation values of the input gate, forget gate, and output gate respectively, c t is the memory cell state, σ(·) represents the Sigmoid activation function, and ⊙ represents the element-wise multiplication operation.

[0039] The trajectory prediction unit is used to predict the future position of the ship based on the output of the time series model. In some embodiments, the prediction result is the trajectory points of the ship at the next n time steps, denoted as The calculation of the prediction points is recursively updated based on the hidden state obtained by forward propagation.

[0040] As an extended technique, this module can also improve the prediction accuracy by fusing historical trajectory data and current environmental information. For example, the model output is corrected by adding the environmental factor weight. Assume that the angle between the directions of the wind speed v w and the water flow speed v f and the ship's running direction is θ, the corrected predicted speed can be expressed as: where k is the correction coefficient and v is the original predicted speed.

[0041] In a possible implementation, the trajectory modeling module can also train the trajectory generation policy using the deep reinforcement learning method to further improve the adaptability to trajectory changes in complex environments. Specifically, the yaw degree, path length, and energy consumption of the generated trajectory are optimized using the reward function.

[0042] Through the above design, the trajectory modeling module can generate the ship's operation trajectory model in an efficient and accurate manner, providing stable basic support for the subsequent anomaly detection and optimization module. This modeling method can still maintain high robustness and real-time performance in complex environments Trajectory anomaly detection module The trajectory anomaly detection module is a key part in this invention to ensure the safety of ship navigation. It is mainly used to analyze the operating trajectory generated by the trajectory modeling module, and identify risks of deviation, collision and other abnormal behaviors. Generally, this module needs to combine real-time trajectory data and environmental information to dynamically evaluate whether the operating state of the ship meets the expectations. The anomaly detection results are directly used as the input for the trajectory optimization module, providing a basis for subsequent path adjustment.

[0043] In this embodiment, the core functions of the trajectory anomaly detection module are jointly realized by the deviation detection unit, the collision risk assessment unit and the abnormal behavior detection unit.

[0044] Specifically, the deviation detection unit is used to monitor in real time whether the ship deviates from the target route. Generally, the target route can be represented as a set of discrete waypoints {P 1 , P 2 , …, P n}, where each point P i includes geographical coordinates (x i , y i ). The deviation distance d 偏 from the current position (x c , y c ) of the ship to the target route can be calculated by the following formula: where (x 1 , y 1 ) and (x 2 , y 2 ) are two adjacent points on the target route. If the deviation distance exceeds the preset threshold d 阈 , it is determined that the ship has deviated.

[0045] As an option, the deviation detection unit can also make a comprehensive judgment by combining the real-time heading angle θ c of the ship and the target heading angle θ t . The angle difference between the two can be calculated by the following formula: Δθ = |θ c - θ t | If the angle difference Δθ exceeds the set value, the deviation behavior is further confirmed.

[0046] The main task of the collision risk assessment unit is to evaluate the potential collision possibility based on the relative position between the current position of the ship and the obstacles. The obstacles can be surrounding ships, floating objects or fixed facilities, etc. Assuming the current position of the ship is x c , y c , and the position of the obstacle is x o , y o , the relative distance d 碰Expressed as: In a possible implementation, the likelihood of a collision can be predicted by introducing a time parameter. Let the speed of the ship be v c and the speed of the obstacle be v o , and the relative speed be v r = v c - v o , and the predicted collision time t 碰 can be expressed as: If t 碰 is less than a preset threshold, a collision alarm is triggered.

[0047] In some embodiments, the collision risk assessment unit further optimizes the risk assessment result in combination with a dynamic danger area model. The dynamic danger area can be represented as an ellipse, the main axis direction of which is consistent with the heading of the ship, and the semi-major axis a and semi-minor axis b are determined by the ship speed and environmental factors. For example: a = k 1 v c + k 2 b = k 3 v c + k 4 where k 1 , k 2 , k 3 , k 4 are model parameters. If the obstacle is within the ellipse, it is determined that there is a collision risk.

[0048] The abnormal behavior detection unit identifies abnormal patterns in the ship's trajectory based on machine learning algorithms. Specifically, normal trajectory patterns are trained through a deep learning model (such as a convolutional neural network CNN or a long short-term memory network LSTM), and behaviors deviating from these patterns are detected. For example, in some embodiments, the spatial features of the trajectory can be extracted through CNN, and the dependencies in the time dimension can be extracted through LSTM. The abnormal score S 异常 can be expressed as: where x 实际 and x 预测 are the actual trajectory point and the predicted trajectory point respectively, and σ is the standard deviation. If S 异常 exceeds the threshold, it is determined as an abnormal behavior.

[0049] As an extended technology, this module also supports retrospective analysis of historical trajectory data. For example, by calculating the average yaw distance and collision risk frequency within a certain time period, it can provide an optimization basis for the path planning module.

[0050] In some embodiments, the abnormal behavior detection unit may also introduce external environmental information (such as wind speed and flow rate) for comprehensive evaluation. For example, if the included angle α between the wind speed v w and the ship's running direction exceeds a certain range, the weight of environmental factors will be introduced into the abnormal score formula: S 环境 = S 异常 + λ·v w ·cosα where λ is an adjustment parameter.

[0051] Through the above design, the trajectory anomaly detection module can effectively identify the risks of yaw, collision and abnormal behavior during ship operation, and provide accurate and timely input data for subsequent trajectory optimization and global trajectory reconstruction. Its multi-level detection mechanism is applicable to complex and changeable navigation environments, and has strong robustness and applicability.

[0052] Trajectory optimization module The trajectory optimization module is one of the key modules in the present invention. It is located after the trajectory anomaly detection module and is mainly used to generate path adjustment suggestions according to the detection results. Generally, the trajectory optimization module needs to comprehensively consider the ship's operation state, surrounding environmental factors and the target route to ensure that the optimized path can avoid risks and improve navigation efficiency. The output of this module is directly used by the global trajectory reconstruction module to provide input for further multi-ship collaboration.

[0053] In this embodiment, the trajectory optimization module is composed of a path calculation unit and an optimization algorithm unit, and combines dynamic programming and multi-objective optimization methods to generate an optimized path that meets the constraint conditions.

[0054] Specifically, the path calculation unit generates a set of path candidates based on the current position and target position of the ship. Generally, the current position (x c , y c ) and the target position (x t , y t ) of the ship are represented as a point pair in a two-dimensional coordinate system, and the candidate paths are composed of multiple discrete points. Each path can be represented as a set of sequence point sets: P = {(x 1 , y 1 ), (x 2 , y 2 ), …, (x n , y n )} where each point (x i , y i ) is a sampling point on the path.

[0055] In a possible implementation, the generation of the path candidate set is based on a rasterization algorithm. Specifically, the navigation area is divided into a number of grid cells, and the center point of each grid cell is used as a sampling point. The path generation process sequentially selects a set of consecutive grid center points from the current position to the target position. To improve the diversity of the path, different sampling densities or step sizes can be set for different candidate paths.

[0056] The optimization algorithm unit evaluates each candidate path according to the candidate set generated by the path calculation unit, in combination with the ship's operating state and environmental factors, and selects the optimal path. As an option, the optimization algorithm unit uses the dynamic programming method to calculate the shortest distance of the path. The state transition formula of dynamic programming is: f(i,j)=min(f(i-1,j-1)+d(i,j),f(i-1,j)+d(i,j),f(i-1,j+1)+d(i,j)) where f(i,j) represents the cumulative shortest distance to the grid point (i,j), and d(i,j) represents the cost of reaching this point.

[0057] In some embodiments, the optimization algorithm unit adopts a multi-objective optimization method, comprehensively considering multiple factors such as path length, collision risk, and navigation energy consumption. The optimization objective can be expressed as: minJ=α·L+β·R+γ·E where L is the total path length, R is the collision risk assessment value of the path, E is the total energy consumption of the path, and α, β, γ are weight parameters.

[0058] The path length L can be expressed as the sum of the distances between all adjacent points on the candidate path: The collision risk assessment value R is calculated in combination with the collision risk results in the trajectory anomaly detection module. The specific formula is: where r i represents the risk value of the i-th point on the path, and this value is related to the distance and relative speed of the obstacle.

[0059] The calculation of the total path energy consumption E can be combined with the ship's speed, flow velocity, and wind speed. The specific formula is: where P i represents the propulsion power at the i-th point, and v i is the ship's speed.

[0060] In a possible implementation, the optimization algorithm unit can also perform a global search through a genetic algorithm (GA) to improve the robustness of the optimization result. The genetic algorithm generates a new population of paths through selection, crossover, and mutation operations, and selects the optimal path based on a fitness function. The definition of the fitness function can be combined with the above multi-objective optimization goals, for example: where J is the multi-objective optimization result.

[0061] As an extended technique, the trajectory optimization module can also consider the impact of real-time environmental changes on the optimized path. For example, when the detected wind speed v w or the flow velocity v f changes significantly, the module will re-optimize the path. Assuming that the angle between the wind direction and the ship's running direction is θ, the correction angle Δφ of the path can be expressed as: Δφ = k·v w ·cosθ where k is an adjustment coefficient.

[0062] In some embodiments, the trajectory optimization module also iteratively updates the optimization strategy by combining historical path data. For example, by analyzing the high-risk points in certain specific areas in historical data, these areas are avoided in advance to generate an optimized path. This optimization method based on historical data can further reduce risks.

[0063] Through the above design, the trajectory optimization module can generate an optimal path that meets various constraint conditions in a complex navigation environment. The path adjustment suggestions output by the module provide a reliable input for the subsequent global track reconstruction module, and at the same time significantly improve the intelligence and real-time performance of path planning Global track reconstruction module The global track reconstruction module is an advanced module in the present invention, mainly used to optimize the overall track distribution of a ship group in a multi-ship cooperation environment. Generally, this module needs to combine the trajectory optimization results of individual ships and the cooperation relationships between multiple ships, and generate a globally coordinated track plan through multi-objective optimization and dynamic reconstruction methods. The global track reconstruction module is a further processing of the output of the trajectory optimization module, and the output result provides a globally coordinated basis for the subsequent feedback and decision support module.

[0064] In this embodiment, the global track reconstruction module is composed of a multi-ship data integration unit, a collaborative analysis unit, and a global path generation unit. The three work together to maximize the overall operating efficiency.

[0065] Specifically, the multi-ship data integration unit is used to collect and integrate the trajectory optimization results and current status information of all ships. Generally, this information includes dynamic data such as the current position, speed, heading, and target position of each ship. In addition, the minimum safe distance between ships, port berth allocation information, and sea condition data also need to be synchronized and integrated.

[0066] As an option, the multi-ship data integration unit standardizes the trajectory optimization results of each ship to ensure that the input data formats for subsequent collaborative analysis are consistent. For example, for the trajectory data of each ship, it can be represented as a set of time-series points: T i ={(t 1 ,x 1 ,y 1 ),(t 2 ,x 2 ,y 2 ),…,(t n ,x n ,y n )} where t k is the time point, and x k ,y k are the corresponding two-dimensional spatial positions.

[0067] The task of the collaborative analysis unit is to construct a global optimization model based on the integrated multi-ship data and dynamically evaluate the collaborative relationships between ships. Generally, the following constraints need to be considered for global optimization: The target position of each ship must be reached within the specified time window; The minimum safe distance between ships shall not be less than the set value; The overall trajectory of the ship group should avoid high-risk areas.

[0068] In a possible implementation, the collaborative analysis unit constructs a global trajectory model based on graph theory. The positions and targets of each ship are regarded as nodes in the graph, and the trajectories between ships are regarded as edges. Assume N i and N j are the positions of two ships, and E ij is the trajectory cost between them. The goal of the model is to minimize the global cost function: where m is the total number of ships, and the cost function E ij can be expressed as a comprehensive weighted value of the relative distance, collision risk, and navigation energy consumption between two ships.

[0069] In some embodiments, the collaborative analysis unit optimizes the global model through a heuristic algorithm (such as the A* algorithm). The algorithm gradually updates the path planning result by evaluating the current state and target position of each ship. For example, the cost evaluation function can be expressed as: f(n) = g(n) + h(n) where g(n) is the actual cost from the starting point to the current node n, and h(n) is the heuristic estimate from the current node to the target node.

[0070] Based on the optimization result of the collaborative analysis unit, the global path generation unit generates the final trajectory of each ship and ensures global consistency. Specifically, this unit needs to dynamically adjust the trajectories of all ships to avoid path intersections or conflicts. In one implementation, the basis for trajectory adjustment is the dynamic interaction area between ships. For example, assume the positions of two ships are (x 1 , y 1 ) and (x 2 , y 2 ), the speeds are v 1 and v 2 , and the relative distance is: If d < d 安全 , then adjust the trajectory direction Δθ: where k is the adjustment coefficient.

[0071] As an extended technique, the global path generation unit can also introduce a dynamic priority mechanism. Specifically, according to the importance of the sailing tasks of the ships or the requirements for the arrival time at the port, different priority weights w i are assigned to each ship. During the global optimization process, the path adjustment range of high-priority ships will be relatively small to ensure the priority completion of their tasks.

[0072] In some embodiments, the module also combines real-time weather and sea condition data to dynamically correct the trajectory generation result. For example, when the wind speed v w or the wave height H in a certain area exceeds the set threshold, the module will adjust the trajectories of all affected ships to avoid entering high-risk areas.

[0073] Through the above design, the global trajectory reconstruction module can generate a globally optimal trajectory distribution in a multi-ship collaborative environment, ensuring the overall safe and efficient operation of the ship group. The flexibility and dynamic adaptability of the module enable it to maintain stable optimization performance in complex sea conditions, and at the same time provide a reliable input basis for the subsequent feedback and decision support module Feedback and Decision Support Module The feedback and decision support module is the final module of the present invention, which is used to transmit the optimization results output by the global track reconstruction module to the user or the automated system of the ship, and assist the decision maker in making operation adjustments. Generally, this module needs to have functions of information visualization, interaction support and operation guidance, so as to intuitively display the optimized path and risk assessment results. In addition, the module also needs to support real-time feedback to ensure the efficient cooperation between the adjustment instructions and the ship operation status. The input of this module mainly comes from the global track reconstruction module, and the output is directly facing the ship control terminal or manual operators.

[0074] In this embodiment, the feedback and decision support module is composed of a result display unit, an interaction control unit and an instruction generation unit.

[0075] Specifically, the result display unit is used to present the optimized track, risk warning information and the overall distribution of the ship group to the user through a graphical interface. Generally, the optimized track can be represented by different colors or line types on a two-dimensional or three-dimensional map, and the current status of each ship is also displayed synchronously. In some embodiments, the track of the ship can be composed of a series of discrete points, and the point set is expressed as: P = {(x 1 , y 1 ), (x 2 , y 2 ), …, (x n , y n )} Among them, each point (x i , y i ) is the geographical coordinate on the map. In order to improve the visualization effect, the line type between points can be dynamically adjusted according to information such as speed and heading.

[0076] As an option, the result display unit can also intuitively present the ship dense area and high-risk area in the form of a heat map. For example, use the risk weight value R i to classify the navigation area: Among them, r ij is the evaluation value of the jth risk factor in the ith area, and w j is its corresponding weight.

[0077] The interaction control unit is used for users to input adjustment instructions and screen and confirm feedback information. Specifically, users can select adjustment paths, avoid specific areas, or prioritize the tasks of a certain ship through this unit. Generally, this unit realizes control functions through buttons, sliders, or gesture operations. In some embodiments, the interaction control unit supports voice recognition, and users can quickly select optimization solutions through voice input. For example, when the user issues the instruction "prioritize arrival at the port", the module will automatically set the current ship as the task with the highest priority.

[0078] In a possible implementation, the interaction control unit also combines multi-mode input, allowing users to synchronously adjust multiple ships. For example, users can select two ships and drag them through the interface to adjust their relative distance. At this time, the system will automatically update the parameter of the relative safety distance d 安全 and recalculate the track.

[0079] The instruction generation unit is responsible for converting the adjustment requirements of users into executable control instructions and sending them to the ship control terminal. Generally, this unit needs to verify the adjustment instructions input by users to ensure that they meet the current navigation conditions. For example, for the instruction to adjust the track, it is necessary to ensure that the adjusted track will not conflict with other ships and will not enter high-risk areas.

[0080] In some embodiments, the instruction generation unit verifies the adjustment results through a simple linear regression model. Suppose the adjusted track is represented as: P′={(x′ 1 ,y 1 ′),(x′ 2 ,y 2 ′),…,(x′ n ,y n ′)} The position of the adjustment point (x i ′,y i ′) needs to meet the following constraints: Among them, d(x i ′,x j ) represents the minimum distance between the adjustment point and the path points of other ships.

[0081] As an extended technology, the instruction generation unit also supports simulating and verifying the adjusted track. For example, the feasibility of the adjustment results is simulated through predefined dynamic environment parameters (such as wind speed and flow rate). Suppose the wind speed in a certain area is v w , the direction is θ w , and the track angle of the adjusted ship is θ c , the module will calculate the influence of the wind direction on the ship: Fw = k·v w ·cos(θ c - θ w ) where k is the wind resistance coefficient. If F w exceeds the set threshold, the instruction generation unit will automatically prompt the user that the adjustment is invalid.

[0082] In some embodiments, the instruction generation unit also supports the generation of multi-ship task collaboration instructions. For example, when multiple ships need to adjust their paths simultaneously to avoid a risk area, the unit will generate a globally consistent adjustment plan by synthesizing the task priorities, current states, and environmental constraints of all ships.

[0083] Through the above design, the feedback and decision support module realizes a complete closed-loop from the optimization result display to the adjustment instruction generation. The flexibility and interactivity of the module provide an intuitive operation experience for the user, while ensuring the accuracy and feasibility of the adjustment instructions. Finally, the output of the module provides strong decision support for the actual operation of the ship, ensuring the safety and efficiency of navigation.

[0084] Please refer to the appendix Figure 2 , as part of this application, the present invention also provides an embodiment, a ship intelligent trajectory analysis method, including the following steps: S1: Collect the basic data of ship operation, environmental information, and multi-source fusion data through the data collection module; Step S1 is the initial link of the method of the present invention, responsible for obtaining the basic data of ship operation, environmental information, and multi-source fusion data through the data collection module. Generally, the execution of this step directly depends on the ship's sensor network, communication equipment, and environmental monitoring system. The goal of step S1 is to generate complete, accurate, and uniformly formatted data, providing reliable input for subsequent trajectory modeling. Logically, this step needs to be closely connected with the data collection module, and its output data will be directly supplied to the trajectory modeling module.

[0085] In this embodiment, the basic data and environmental information of the ship are dynamically obtained through the data collection module.

[0086] Specifically, the basic data includes operation parameters such as the speed, position, heading, and acceleration of the ship. In some embodiments, the position data is obtained in real time through an on-board GPS system and represented in the form of geographical coordinates (x, y, z), where x and y are the ground plane coordinates and z is the height. The speed information is calculated by an inertial navigation system (INS) and is usually represented in vector form as, v x , v y , v z being the components in three directions.

[0087] The above steps achieve a comprehensive perception of the ship's operating status and the surrounding environment, and generate multi-dimensional fusion data, providing complete and high-quality input for subsequent trajectory modeling. Its efficient data acquisition and fusion capabilities are applicable to complex and dynamic navigation scenarios.

[0088] S2: Establish a ship operation trajectory model through the trajectory modeling module; Step S2 is one of the core links of the method of the present invention, mainly used to transmit the data collected and fused in step S1 to the data storage module and the ship intelligent management system. Generally, this step needs to ensure the stability and timeliness of data transmission, while ensuring the integrity and accuracy of the transmitted data. In the overall system architecture, step S2 is directly connected to the data acquisition module, the data storage module, and the ship intelligent management system, providing real-time and reliable data support for subsequent trajectory modeling and analysis.

[0089] In this embodiment, step S2 realizes the efficient transmission of data through the network communication module, specifically including data distribution, communication protocol management, and data integrity verification.

[0090] Specifically, data distribution is one of the core tasks of step S2, responsible for distributing the multi-dimensional fusion data output by step S1 to different modules according to functional requirements. In some embodiments, data distribution is based on task priorities, and key data is preferentially transmitted to the ship intelligent management system. For example, basic data such as the real-time position, speed, and heading of the ship can be directly used by the trajectory modeling module, while environmental information (such as wind speed and wave height) is stored in the data storage module for subsequent analysis.

[0091] When the system detects an abnormal or high-risk event (such as a yaw or collision risk), the module will preferentially transmit data related to the risk. For example, the relative position data (x c , y c ) and (x o , y o ) between the ship and the obstacle can calculate the relative distance d through the following formula: If d is less than the preset safety distance d 安全 , the relevant data will be preferentially marked and quickly transmitted.

[0092] Communication protocol management is used to standardize the format and path of data during transmission to ensure compatibility between modules. Data integrity verification is one of the important technical steps in step S2 and is used to ensure that data is not lost or tampered with during transmission. Generally, integrity verification is achieved through a checksum (such as CRC verification). In some embodiments, the module generates a checksum for each data segment before data distribution and recalculates the checksum after the data reaches the target module for comparison. If the checksums do not match, data retransmission is triggered.

[0093] Data integrity verification can also combine a fragmentation verification mechanism to perform block verification on large data packets.

[0094] At the same time, the data transmission in step S2 also supports a multi-node transmission mechanism to improve the reliability and redundancy of transmission. For example, when the main communication channel fails, the system will automatically switch to the backup channel for data transmission. At this time, timestamp synchronization of the data is particularly crucial to ensure that the data received by different nodes has a consistent time stamp.

[0095] Through the above design, step S2 realizes an efficient transmission process of data from collection to storage and use, providing complete and real-time data support for the trajectory modeling module and subsequent steps. While taking into account transmission stability and efficiency, this step further ensures data reliability through various verification and redundancy mechanisms. S3: Detect yaw, collision risks, or other abnormal behaviors in the trajectory through the trajectory anomaly detection module; Step S3 is an important part of the present invention and is mainly used to input the multi-dimensional fusion data stored in the data storage module into the trajectory modeling module and complete the trajectory analysis process. Generally, this step needs to make full use of real-time data and historical data, through time series analysis and modeling, to predict the future operating trajectory of the ship and at the same time provide accurate input data for subsequent modules. The output result of step S3 is directly connected to the trajectory anomaly detection module and is used to identify potential navigation anomalies and risks.

[0096] In this embodiment, step S3 includes three main processes: data preprocessing, trajectory modeling, and trajectory prediction, which are implemented by combining deep learning methods such as recurrent neural networks (RNN) and long short-term memory networks (LSTM).

[0097] Specifically, data preprocessing is the basic link of trajectory modeling and is mainly used to clean, normalize, and time-align the input data. In some embodiments, data preprocessing first processes missing values and outliers. Generally, missing values can be filled in by interpolation.

[0098] Trajectory modeling is the core task of step S3, which is responsible for establishing a time series model based on the preprocessed input data to capture the changing patterns of the ship's motion state. Generally, this process is implemented using a Recurrent Neural Network (RNN). The advantage of RNN lies in its ability to handle sequence data with time dependence, and the update formula for its hidden state is: h t =f(W h h t-1 +W x x t +b h where h t represents the hidden state at the t-th moment, W h and W x are the weight matrices of the hidden layer and the input layer respectively, b h is the bias vector, and f(·) is the activation function.

[0099] To address the problem of vanishing gradients in RNN for processing long sequences, in some embodiments, the Long Short-Term Memory network (LSTM) is introduced. LSTM introduces memory cells in the network structure and selectively updates information through input gates, forget gates, and output gates.

[0100] Trajectory prediction is used to generate future trajectory points of the ship based on the results of trajectory modeling, usually for short-term prediction.

[0101] Through the above design, step S3 completes the whole process from data modeling to trajectory prediction, providing accurate input data for the trajectory anomaly detection and path optimization modules. By combining deep learning techniques, this step has significant robustness and adaptability in dealing with complex trajectory patterns and dynamic environment data S4: Generate path adjustment suggestions through the trajectory optimization module; Step S4 is one of the important functional steps of the present invention, which is used to generate targeted path adjustment suggestions according to the output results of the trajectory anomaly detection module. Generally, this step needs to comprehensively consider the ship's current position, target route, environmental conditions, and detected anomaly features to formulate an optimized path. The path optimization result is directly used to guide the real-time adjustment of the ship and provides input for the subsequent global track reconstruction module.

[0102] In this embodiment, step S4 includes three main processes: risk assessment, path optimization, and result output, which are implemented by combining dynamic programming and multi-objective optimization methods.

[0103] Specifically, risk assessment is a precondition for path optimization and is used to quantify the impact of detected anomalies on path safety. In some embodiments, the main inputs for risk assessment include the relative position, relative speed between the ship and obstacles, and environmental factors. For example, assume the current position of the ship is (x c , y c ), the position of the obstacle is (x o , y o ), and the relative distance d between the two can be expressed as: If the distance d is less than the preset safety threshold d 安全 , it is determined that there is a collision risk. The calculation of the risk value R can be further combined with the relative speed v r and the relative direction angle θ r : where k 1 , k 2 are adjustment coefficients, and θ r is the relative direction angle.

[0104] Path optimization is the core part of step S4 and is used to generate the optimal path that meets the current navigation conditions. Generally, this process is implemented based on the dynamic programming method, and the optimal solution on the path is found through recursive calculation.

[0105] Through the above design, step S4 takes into account both risk avoidance and navigation efficiency when generating path adjustment suggestions, providing an intelligent solution for the dynamic adjustment of the ship. By combining the multi-objective optimization method and the dynamic correction mechanism, this step can adapt to the complex and changeable navigation environment and at the same time provide an accurate input basis for the subsequent global track reconstruction.

[0106] S5: Based on multi-ship data, the global track reconstruction module performs collaborative analysis and reconstructs the optimized route; Step S5 is the global optimization part of the present invention and is mainly used to perform collaborative analysis in a multi-ship operation environment, reconstruct the global track, and optimize the overall operation efficiency and safety of the ship group. Generally, this step needs to combine the current states, target paths, and environmental data of all ships, and generate a globally consistent track plan through collaborative calculation. Step S5 is closely connected with the path optimization module and is an extension and integration of path adjustment, aiming to improve the effect of multi-ship collaborative operation.

[0107] In this embodiment, step S5 includes three core processes: data integration, collaborative analysis, and global path generation.

[0108] Specifically, data integration is the starting point of global track reconstruction, which is used to summarize the trajectory optimization results of all ships and the collaborative analysis of relevant operation parameters is the core task of step S5. It is responsible for building a global track optimization model based on the integrated data and coordinating the relative positions and path distributions among multiple ships. Generally, the collaborative analysis needs to comprehensively consider the following constraint conditions: The minimum safe distance between ships; The target position and arrival time of each ship; The influence of environmental factors on the path.

[0109] Global path generation is the output link of collaborative analysis, which is responsible for allocating the optimized global path to each ship and ensuring the coordination and consistency among the paths.

[0110] Through the above design, step S5 realizes the global optimization of multi-ship operation and generates a coordinated track distribution. While taking into account path safety and operation efficiency, the module also supports a real-time adjustment mechanism, can adapt to complex and dynamic navigation environments, and provides comprehensive and reliable input for subsequent decision-making support.

[0111] S6: Feed back the track optimization results to the ship intelligent management system through the feedback and decision-making support module for display and decision adjustment.

[0112] Step S6 is the final step of the method of the present invention, which is responsible for feeding back the optimization results generated by the global track reconstruction module to the ship intelligent management system and providing necessary decision-making support information at the same time. Generally, step S6 needs to comprehensively display the optimized path, risk assessment results and real-time adjustment suggestions to provide intuitive and operable guidance for the operation of the ship. Closely connected with the previous steps, this step is the key link to ensure the actual implementation of the optimization plan.

[0113] In this embodiment, step S6 includes three main processes: information feedback, user interaction and decision-making support, to ensure that the optimization results can be efficiently transmitted and guide ship operation.

[0114] Specifically, information feedback is the core task of step S6, which is used to display the results of global track reconstruction to the ship intelligent management system in a multi-dimensional form. In some embodiments, the optimized path is displayed on the electronic nautical chart in a two-dimensional or three-dimensional graphical interface. The path point set of each ship can be represented by discrete data as: P = {(x 1 , y 1 ), (x 2 , y 2 ), …, (x n , y n )} where each point (x i , y i) are coordinates on the map. The path display uses colors or line types to distinguish high-risk areas from safe areas. For example, a red path can be used to mark high-risk sections, while a green path indicates a recommended safe path.

[0115] User interaction is a key part of step S6 and is used to support manual intervention and dynamic adjustment. Under normal circumstances, the user can select different path adjustment options through the interface. For example, the user may adjust the path of a certain ship according to weather changes or special shipping tasks.

[0116] Decision support is the core output part of step S6 and is used to convert the optimization results and user adjustments into actual operation instructions. In some embodiments, the decision support module combines the current environmental data and the path optimization results to generate specific operation suggestions. For example, when the angle θ between the wind speed direction and the ship's heading exceeds a set threshold, the module will suggest adjusting the heading φ′, and the calculation formula is: φ′ = φ + k·cos(θ) where φ is the original heading and k is the adjustment coefficient.

[0117] Through the above design, step S6 realizes a complete closed loop from the optimization result feedback to decision support. While providing clear and intuitive information, the module supports flexible adjustment by the user to ensure that the optimized path can adapt to the changes in the dynamic environment. The finally output decision support instructions not only improve the safety and efficiency of operation but also provide greater flexibility and autonomy for ship operation.

[0118] Although the embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A ship intelligent trajectory analysis system, characterized in that: include, Data acquisition module, used to collect basic ship operation data, environmental information and multi-source fusion data; A trajectory modeling module, used to establish a ship operation trajectory model based on the multi-dimensional data acquired by the data acquisition module; The trajectory anomaly detection module is used to identify deviation, collision risk and abnormal behavior based on the trajectory model generated by the trajectory modeling module; The trajectory optimization module is used to generate path adjustment suggestions based on the detection results of the trajectory anomaly detection module; The global track reconstruction module is used to reconstruct and optimize the overall track of the ship group in combination with cluster collaborative analysis; The feedback and decision support module is used to feed back trajectory analysis results and optimization suggestions to the ship intelligent management system to support route adjustment decisions.

2. A ship intelligent trajectory analysis system according to claim 1, characterized in that: The data acquisition module comprises: Ship operation status collection unit, used to collect ship speed, heading and position data; Environmental data acquisition unit, used to obtain weather information, sea condition data and obstacle information; The data fusion unit is used to integrate multi-source data and output them in a unified data format.

3. A ship intelligent trajectory analysis system according to claim 1, characterized in that: The trajectory modeling module predicts the ship trajectory based on a time series algorithm, and the prediction process includes: Use recurrent neural networks to perform time series analysis on ship operating status and environmental data; Predict the next location of the ship based on its historical trajectory and current status.

4. A ship intelligent trajectory analysis system according to claim 1, characterized in that: The trajectory anomaly detection module includes: A deviation detection unit is used to calculate the deviation between the current position of the ship and the target route; A collision risk assessment unit, used to calculate a collision risk value based on the distance and relative speed between the current position of the ship and the obstacle; Abnormal behavior detection unit, used to identify abnormal trajectory patterns through deep learning models.

5. A ship intelligent trajectory analysis system according to claim 1, characterized in that: The trajectory anomaly detection module includes: A deviation detection unit is used to calculate the deviation between the current position of the ship and the target route; A collision risk assessment unit, used to calculate a collision risk value based on the distance and relative speed between the current position of the ship and the obstacle; Abnormal behavior detection unit, used to identify abnormal trajectory patterns through deep learning models.

6. A ship intelligent trajectory analysis method, characterized in that: A ship intelligent trajectory analysis system applied to any one of claims 1 to 5, comprising the following steps: S1: Collect basic ship operation data, environmental information and multi-source fusion data through the data acquisition module; S2: Establishing the ship operation trajectory model through the trajectory modeling module; S3: Detect the yaw and collision risks in the trajectory through the trajectory anomaly detection module; S4: Generate path adjustment suggestions through trajectory optimization module; S5: The global track reconstruction module is used to perform collaborative analysis based on multi-ship data and reconstruct the optimized route; S6: The trajectory optimization results are fed back to the ship intelligent management system through the feedback and decision support module for display and adjustment of decisions.

7. A ship intelligent trajectory analysis method according to claim 6, characterized in that: The trajectory modeling in step S2 includes: Establish trajectory prediction model based on time series data of ship operation; The trajectory model is updated in real time through the long short-term memory network; Output the predicted track points and status information of the ship.

8. A ship intelligent trajectory analysis method according to claim 6, characterized in that: The path optimization in step S4 includes the following steps: Calculate the risk weight based on the detection results of the trajectory anomaly detection module; Use dynamic programming algorithm to generate the shortest path and the lowest risk path; Output the optimized path for reference by the ship intelligent management system.

9. A ship intelligent trajectory analysis method according to claim 6, characterized in that: The global track reconstruction in step S5 comprises the following steps: Establish a global collaborative model based on multi-ship data; Calculate the optimal distribution of all ships in the target area; Dynamically adjust the routes of each ship to maximize overall operating efficiency.

10. A ship intelligent trajectory analysis method according to claim 6, characterized in that: The feedback and decision support in step S6 include: Display trajectory optimization results and risk warning information through a graphical interface; Provide an operation command interface so that users can adjust the ship's operation strategy.

Images