Ship trajectory similarity judgment method and system based on multiple dimensions and dynamic weights

Abstract

The invention relates to the technical field of ship intelligent navigation and trajectory analysis, and discloses a ship trajectory similarity judgment method and system based on multiple dimensions and dynamic weights, and the method comprises the steps: obtaining load data, real-time navigation parameters and environmental parameters, and generating a load state classification result; calculating according to the load state classification result to obtain an acceleration performance index and a turning radius change trend; establishing a mathematical relationship model according to the load state classification result and the turning radius change trend; generating a dynamic incidence matrix according to the mathematical relationship model and the acceleration performance index; according to the dynamic incidence matrix, dynamically adjusting a weight value of a load state to obtain a weight distribution result; and performing calculation according to the weight distribution result and the real-time navigation parameters to obtain a final similarity judgment result. According to the method, the dynamic association of the ship load state and the trajectory characteristics can be realized, and the weight self-adaptive adjustment is realized in the similarity judgment process.

Description

Technical Field

[0001] The present invention relates to the technical field of ship intelligent navigation and trajectory analysis, and in particular to a method and system for judging ship trajectory similarity based on multi-dimensionality and dynamic weights. Background Art

[0002] Currently, intelligent ship navigation and trajectory analysis technologies face the challenges of a complex and ever-changing navigation environment. With the continued growth of global shipping density and the trend toward larger ships, traditional trajectory analysis methods are increasingly limited in handling dynamic scenarios such as fluctuating ship loads and sudden maneuvers. Real-time changes in a ship's load state directly affect its inertial characteristics and maneuverability, making it difficult for trajectory matching methods based on fixed parameters to accurately assess a ship's true motion intentions. This is especially true in high-risk scenarios such as typhoons and congested waterways. Static weight allocation mechanisms can easily lead to misjudgments of trajectory similarity, seriously impacting the timeliness and safety of collision avoidance decisions.

[0003] In one existing technology, ship trajectory similarity determination primarily relies on a geometric feature space matching method. Trajectory comparisons are performed using preset speed weights and heading angle thresholds, a sliding window mechanism is used to extract the spatial distribution characteristics of trajectory points, and a fixed-ratio environmental parameter correction coefficient is used to compensate for wind and wave interference. While this method can address basic position offset issues, it treats the ship as a rigid body model with constant mass, ignoring changes in the ship's hydrodynamic characteristics caused by changes in load, which directly leads to significant deviations between the trajectory arc curvature and theoretical predictions. Furthermore, the existing weight allocation mechanism uses an offline training model and is unable to adjust the calculation strategy based on dynamic factors such as real-time collected load data and sudden avoidance instructions. When encountering sudden operating conditions such as cargo shifting and ballast water adjustment, the fixed weight coefficient cannot accurately reflect the correlation strength between the load state and trajectory characteristics, ultimately resulting in a break in the mapping relationship between load characteristics and motion trajectory.

[0004] Therefore, the existing technology has the problem that it is difficult to dynamically associate the ship load status with the trajectory characteristics, and the weight adaptive adjustment is not achieved in the similarity judgment process. Summary of the Invention

[0005] The present invention provides a method and system for judging the similarity of ship trajectories based on multi-dimensionality and dynamic weights, so as to dynamically associate the ship load status with the trajectory characteristics and realize adaptive weight adjustment in the similarity judgment process.

[0006] In a first aspect, in order to solve the above technical problems, the present invention provides a method for determining the similarity of ship trajectories based on multi-dimensional and dynamic weights, comprising:

[0007] Obtain load data, real-time navigation parameters, and environmental parameters from the ship's navigation data, analyze them based on the reward mechanism, and generate load status classification results;

[0008] Calculating acceleration characteristics and performing turning radius regression analysis based on the load state classification result, the real-time navigation parameters, and the environmental parameters to obtain acceleration performance indicators and turning radius change trends under different load states;

[0009] Analyzing the dynamic behavior of the ship based on the load state classification results and the turning radius change trend, and establishing a mathematical relationship model between the load state and the trajectory arc characteristics;

[0010] According to the mathematical relationship model between the load state and the trajectory arc characteristics, the acceleration performance index, the real-time navigation parameters and the environmental parameters, the dynamic course adjustment characteristics of the dense navigation area are analyzed, and correlation modeling is performed to generate a dynamic correlation matrix between the load state and the trajectory characteristics;

[0011] According to the dynamic correlation matrix, the real-time navigation parameters and the environmental parameters, the trajectory similarity is dynamically adjusted to calculate the weight value of the load state and obtain a weight distribution result;

[0012] A cluster analysis is performed based on the weight distribution result and the real-time navigation parameters. The similarity of trajectories with similar load states is calculated in combination with preset path constraints to obtain a final similarity judgment result.

[0013] In an optional embodiment, the acquiring of load data, real-time navigation parameters, and environmental parameters from the ship navigation data, and analyzing them based on a reward mechanism to generate a load status classification result includes:

[0014] Obtain load data, real-time navigation parameters and environmental parameters from ship navigation data;

[0015] According to the load data, the real-time navigation parameters and the environmental parameters, a preset reward mechanism is used to calculate a reward value; wherein the size of the reward value reflects the quality of the ship's operating status;

[0016] Performing preliminary classification using a CART algorithm based on the load data, the real-time navigation parameters, the environmental parameters, and the reward value to obtain a classification feature vector containing a ship operation status value;

[0017] According to the classification feature vector, a support vector machine algorithm is used to perform vector optimization to obtain a modified classification boundary condition;

[0018] A multi-dimensional cluster analysis is performed based on the modified classification boundary conditions, the classification feature vectors, and the reward value to obtain a load state classification result including normal load, overload, and optimal load states.

[0019] In an optional embodiment, the calculation of acceleration characteristics and the performance of turning radius regression analysis based on the load state classification result, the real-time navigation parameters, and the environmental parameters to obtain acceleration performance indicators and turning radius change trends under different load states includes:

[0020] Performing classification calculations based on the load state classification results, the real-time navigation parameters, and the environmental parameters to obtain ship mass distribution parameters and hydrodynamic resistance coefficients;

[0021] Calculating the ship acceleration and acceleration response time based on the real-time navigation parameters and the ship mass distribution parameters, performing time series analysis on the obtained ship acceleration and acceleration response time, and obtaining acceleration performance indicators under different load conditions;

[0022] A regression analysis is performed based on the ship mass distribution parameters, the hydrodynamic resistance coefficient, the environmental parameters and the real-time navigation parameters to infer a mapping relationship between the load state and the turning radius change rate, thereby obtaining a turning radius change trend.

[0023] In an optional embodiment, analyzing the dynamic behavior of the ship based on the load state classification result and the turning radius change trend, and establishing a mathematical relationship model between the load state and the trajectory arc characteristics, includes:

[0024] Performing a time series analysis based on the load state classification results and the turning radius variation trend to obtain dynamic correlation features; performing a correlation analysis using a random forest algorithm based on the dynamic correlation features, the load state classification results, and the turning radius variation trend to obtain a core parameter combination that affects trajectory arc features;

[0025] The core parameter combination includes load state, track curvature and steering angle rate;

[0026] According to the core parameter combination, a multivariate regression equation is constructed, and the nonlinear relationship between the load state and the track curvature and the steering angle rate is fitted using the least squares method to obtain a load-track regression model;

[0027] According to the load-track regression model, the load state classification result and the turning radius change trend, combined with the standard trajectory arc parameters in a preset ship maneuvering feature library, the load-track regression model is constrainedly optimized to obtain a mathematical relationship model between load state and trajectory arc characteristics.

[0028] In an optional embodiment, the dynamic course adjustment characteristics of the dense navigation area are analyzed based on the mathematical relationship model between the load state and the trajectory arc characteristics, the acceleration performance index, the real-time navigation parameters, and the environmental parameters, and correlation modeling is performed to generate a dynamic correlation matrix between the load state and the trajectory characteristics, including:

[0029] Performing a multivariate collaborative analysis based on the mathematical relationship model between the load state and the trajectory arc characteristics, the real-time navigation parameters, and the environmental parameters to obtain a heading adjustment response speed and a heading adjustment amplitude;

[0030] Calculate the dynamic maneuvering margin of the ship in a dense navigation area based on the acceleration performance index and the real-time navigation parameters, and construct a course adjustment urgency assessment index based on a preset collision avoidance safety distance threshold;

[0031] Based on the mathematical relationship model between the load state and the trajectory arc characteristics, the heading adjustment response speed and the heading adjustment amplitude, a convolutional neural network is used to perform spatiotemporal correlation analysis to identify the implicit influence of the load state on the heading adjustment mode;

[0032] Based on the implicit influence relationship, the acceleration performance index, the real-time navigation parameters and the environmental parameters, multi-scale matching is performed using multiple linear regression to establish a three-dimensional mapping relationship between load state, acceleration performance and heading adjustment sensitivity;

[0033] According to the three-dimensional mapping relationship and the heading adjustment urgency evaluation index, K-means cluster analysis is used to perform correlation modeling to obtain a dynamic correlation matrix between load status and trajectory characteristics.

[0034] In an optional embodiment, dynamically adjusting the weight value of the load state in the trajectory similarity calculation according to the dynamic correlation matrix, the real-time navigation parameters, and the environmental parameters to obtain a weight distribution result includes:

[0035] Performing adaptive weight allocation according to the dynamic correlation matrix to construct a weight allocation feature space, including load state weight, heading adjustment priority, and trajectory deviation tolerance;

[0036] The load state weight, the real-time navigation parameters and the environmental parameters are used to establish a weight distribution strategy using a Bayesian optimization algorithm, and dynamically adjust the basic distribution ratio of the load state weight to obtain a first load state weight;

[0037] According to the course adjustment priority, the real-time navigation parameters and the environmental parameters, combined with the safe operation threshold in the preset ship collision avoidance rule library, the course adjustment priority is normalized to generate a dynamic constraint condition for weight adjustment;

[0038] Based on the trajectory deviation tolerance and the first load state weight, a reinforcement learning algorithm is used to perform real-time learning. When a risk of crossing the trajectories of adjacent ships is detected, the load state weight is increased in real time to obtain a second load state weight.

[0039] Based on the second load state weight, the dynamic constraints, the real-time navigation parameters, and the environmental parameters, trajectory similarity calculation is performed. When the trajectory similarity is less than a preset trajectory similarity threshold, the load state weight is dynamically corrected in combination with the dynamic constraints to obtain a weight distribution result. When the trajectory similarity is greater than the preset trajectory similarity threshold, the second load state weight is determined as the final load state weight to obtain a weight distribution result. In an optional embodiment, cluster analysis is performed based on the weight distribution result and the real-time navigation parameters, and a dynamic time warping algorithm is used to calculate the similarity of load state-similar trajectories in combination with path constraints to obtain a final similarity judgment result, including:

[0040] Extracting navigation track features based on the weight distribution result and the real-time navigation parameters, performing track standardization processing, and generating a track feature vector data set;

[0041] Based on the trajectory feature vector dataset, a density clustering algorithm is used to perform spatial distribution analysis, and classification is performed according to the load state weight ratio to obtain trajectory clusters including high similarity, medium similarity, and low similarity;

[0042] Performing dynamic time warping on the trajectory clusters, and combining the minimum turning radius threshold and the safe speed limit in the preset path constraints to calculate the dynamic bending distance between the trajectory sequences;

[0043] Perform risk compensation analysis based on the dynamic bending distance and the weight distribution result, construct a multi-dimensional similarity evaluation index, and perform similarity sorting on the trajectories in the same cluster to obtain a similarity sorting result;

[0044] According to the similarity sorting result and the real-time navigation parameters, time and space alignment is performed to obtain a final similarity judgment result including load state matching degree, path consistency coefficient and risk warning level.

[0045] In a second aspect, the present invention provides a device for determining the similarity of ship trajectories based on multi-dimensionality and dynamic weights, comprising:

[0046] The data acquisition module is used to obtain load data, real-time navigation parameters and environmental parameters from the ship's navigation data, analyze them based on the reward mechanism, and generate load status classification results;

[0047] a speed-path analysis module, configured to calculate acceleration characteristics and perform turning radius regression analysis based on the load state classification result, the real-time navigation parameters, and the environmental parameters, to obtain acceleration performance indicators and turning radius change trends under different load states;

[0048] A regression modeling module is used to analyze the dynamic behavior of the ship based on the load state classification results and the turning radius change trend, and to establish a mathematical relationship model between the load state and the trajectory arc characteristics;

[0049] a dynamic matrix module for analyzing the dynamic course adjustment characteristics of the dense navigation area based on the mathematical relationship model between the load state and the trajectory arc characteristics, the acceleration performance index, the real-time navigation parameters, and the environmental parameters, performing correlation modeling, and generating a dynamic correlation matrix between the load state and the trajectory characteristics;

[0050] a weight distribution module, configured to dynamically adjust the weight value of the load state in the trajectory similarity calculation according to the dynamic association matrix, the real-time navigation parameters, and the environmental parameters to obtain a weight distribution result;

[0051] The result output module is used to perform cluster analysis based on the weight distribution result and the real-time navigation parameters, and calculate the similarity of similar load state trajectories in combination with preset path constraints to obtain a final similarity judgment result.

[0052] In a third aspect, the present invention also provides an electronic device comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, the method for determining the similarity of ship trajectories based on multi-dimensionality and dynamic weights described above is implemented.

[0053] In a fourth aspect, the present invention also provides a computer-readable storage medium, which includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute any one of the above-mentioned ship trajectory similarity judgment methods based on multi-dimensionality and dynamic weights.

[0054] Compared with the prior art, the present invention has the following beneficial effects:

[0055] (1) This invention obtains load data, real-time navigation parameters, and environmental parameters from ship navigation data, generates load status classification results based on reward mechanism analysis, and improves classification accuracy and efficiency. It simplifies the complexity of initial dynamic physical inspection parameters and reduces the amount of subsequent calculations through CART algorithm and support vector machine optimization. Based on the reward mechanism and multi-dimensional clustering, the classification results are more consistent with the actual operating status, including normal, overloaded, and optimal load states, providing an accurate basis for subsequent analysis, reducing classification bias, and improving the accuracy of trajectory similarity judgment.

[0056] (2) Based on the load state classification results, real-time navigation parameters, and environmental parameters, the present invention calculates acceleration characteristics and performs turning radius regression analysis to obtain acceleration performance indicators and turning radius change trends, thereby improving the efficiency and accuracy of dynamic behavior analysis. The ship's mass distribution parameters and hydrodynamic resistance coefficient are classified and calculated. The acceleration and acceleration response time are calculated in combination with real-time navigation parameters, and time series analysis is performed, reducing the computational complexity caused by data redundancy. The mapping relationship between load state and turning radius change rate, derived through regression analysis, provides accurate data for subsequent modeling.

[0057] (3) Based on the load state classification results and the trend of turning radius changes, the present invention analyzes the dynamic behavior of the ship and establishes a mathematical relationship model between the load state and the trajectory arc characteristics. The random forest algorithm selects core parameters, reduces the interference of irrelevant parameters, and improves computational efficiency. Through the least squares method to fit the nonlinear relationship and constrained optimization, the mathematical relationship model between the load state and the trajectory arc characteristics is more consistent with the dynamic behavior, reducing model errors and providing a reliable basis for subsequent heading adjustment analysis.

[0058] (4) Based on mathematical relationship models, acceleration performance indicators, real-time navigation parameters, and environmental parameters, this invention analyzes the dynamic course adjustment characteristics of dense navigation areas and generates a dynamic correlation matrix, thereby improving the efficiency and accuracy of trajectory feature analysis. A dynamic correlation matrix of load status and trajectory characteristics is obtained through multiple linear regression and K-means clustering, providing precise support for subsequent weight allocation, avoiding analysis bias, and improving the accuracy of trajectory similarity calculation.

[0059] (5) The present invention dynamically adjusts the weight value of the load state in the trajectory similarity calculation based on the dynamic association matrix, real-time navigation parameters and environmental parameters to obtain a weight distribution result, thereby improving the efficiency and accuracy of weight distribution. By constructing a weight distribution feature space through adaptive weight distribution, the load state weight is dynamically adjusted in combination with the Bayesian optimization algorithm and the reinforcement learning algorithm, reducing the calculation deviation caused by fixed weights and improving the efficiency of weight adjustment. By normalizing the course adjustment priority and combining it with the ship collision avoidance rule library to generate dynamic constraint conditions, the weight is comprehensively corrected, and the generated weight distribution result is more in line with the actual navigation scene, avoiding the similarity calculation error caused by unreasonable weight distribution.

[0060] (6) The present invention performs cluster analysis based on the weight distribution results and real-time navigation parameters, and calculates the similarity of similar load state trajectories in combination with preset path constraints to obtain the final similarity judgment result, thereby improving the efficiency and accuracy of trajectory similarity calculation. By extracting navigation trajectory features and performing normalization processing to generate a trajectory feature vector data set, a density clustering algorithm is used to classify trajectory clusters, reducing the computational complexity caused by uneven data distribution; combining path constraints to calculate dynamic bending distance, and constructing a multidimensional similarity evaluation index through risk compensation analysis to generate a similarity ranking result, thus avoiding the deviation caused by misalignment of time series; finally, through spatiotemporal alignment, a final similarity judgment result including load state matching degree, path consistency coefficient and risk warning level is generated, providing a reliable basis for navigation optimization and safety warning. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 1 is a flow chart of a method for determining ship trajectory similarity based on multi-dimensionality and dynamic weights provided by the first embodiment of the present invention;

[0062] Figure 2 2 is a schematic diagram of the structure of a ship trajectory similarity judgment system based on multi-dimensionality and dynamic weights provided in the second embodiment of the present invention. DETAILED DESCRIPTION

[0063] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0064] Reference Figure 1 The first embodiment of the present invention provides a method for determining the similarity of ship trajectories based on multi-dimensionality and dynamic weights, comprising the following steps:

[0065] S11, obtaining load data, real-time navigation parameters and environmental parameters from the ship's navigation data, analyzing them based on the reward mechanism, and generating a load status classification result;

[0066] S12, calculating acceleration characteristics and performing turning radius regression analysis based on the load state classification result, the real-time navigation parameters, and the environmental parameters to obtain acceleration performance indicators and turning radius change trends under different load states;

[0067] S13, analyzing the dynamic behavior of the ship based on the load state classification result and the turning radius change trend, and establishing a mathematical relationship model between the load state and the trajectory arc characteristics;

[0068] S14, analyzing the dynamic course adjustment characteristics of the dense navigation area based on the mathematical relationship model between the load state and the trajectory arc characteristics, the acceleration performance index, the real-time navigation parameters, and the environmental parameters, performing correlation modeling, and generating a dynamic correlation matrix between the load state and the trajectory characteristics;

[0069] S15, dynamically adjusting the weight value of the load state in the trajectory similarity calculation according to the dynamic correlation matrix, the real-time navigation parameters, and the environmental parameters to obtain a weight distribution result;

[0070] S16, performing cluster analysis based on the weight distribution result and the real-time navigation parameters, and calculating the similarity of trajectories with similar load states in combination with preset path constraints to obtain a final similarity judgment result.

[0071] In step S11, it is necessary to obtain the load data, real-time navigation parameters and environmental parameters in the ship navigation data, analyze them based on the reward mechanism, and generate a load status classification result.

[0072] In one implementation, load data, real-time navigation parameters, and environmental parameters are obtained from the ship's navigation data, and analyzed based on a reward mechanism to generate a load status classification result, including:

[0073] Obtain load data, real-time navigation parameters, and environmental parameters from the ship's navigation data; calculate a reward value using a preset reward mechanism based on the load data, the real-time navigation parameters, and the environmental parameters; perform preliminary classification using a CART algorithm based on the load data, the real-time navigation parameters, the environmental parameters, and the reward value to obtain a classification feature vector containing a ship's operating status value; perform vector optimization using a support vector machine algorithm based on the classification feature vector to obtain a revised classification boundary condition; perform a multidimensional cluster analysis based on the revised classification boundary condition, the classification feature vector, and the reward value to obtain a load state classification result including normal load, overload, and optimal load states.

[0074] It should be noted that, based on the load data, the real-time navigation parameters and the environmental parameters, a preset reward mechanism is used to calculate the reward value, wherein the preset reward mechanism first sets the evaluation criteria according to the ship operation target, for example, with the speed stability and load rationality as the target, the load data is compared with the standard load range, and the operation efficiency is calculated by combining the speed and water flow speed. The reward value is generated by weighted calculation based on these factors. The size of the reward value reflects the quality of the ship's operation status and is used to quantify the ship's operation performance in subsequent classification. The reward value R is composed of the load rationality score (W score ) and speed stability score (V score ) weighted calculation, and the introduction of the operating efficiency coefficient (η) correction, the final formula is:

[0075] R=α·W score +β·V score ·η

[0076] Where α+β=1 (α, β are preset weight coefficients, the default values ​​are α=0.4, β=0.6), and η∈0,1] is the operating efficiency coefficient.

[0077] Loading rationality score (W score ) is calculated as:

[0078]

[0079] Among them, the standard load range of ships is [W min ,W max ],W min ,W max They are the maximum and minimum values ​​of the preset standard load of the ship, and the actual load is W actual .

[0080] Speed ​​stability score (V score ): According to the performance setting of the ship, the theoretical optimal speed V is obtained target According to the speed fluctuation, the actual speed sequence within the time window t is taken Calculate the standard deviation using the formula:

[0081]

[0082] The formula for calculating the speed stability score based on the standard deviation is:

[0083]

[0084] Where e is the base of the natural logarithm, k is the decay coefficient, and k = 0.5. (Score range: 0-1, the smaller the fluctuation, the higher the score)

[0085] The operating efficiency coefficient (η) is calculated using the following formula:

[0086] V ground =V actual +C current cosθ

[0087]

[0088] Among them, V actual is the actual sailing speed of the ship, C current is the water velocity obtained by measurement, V ground is the actual ground speed, and θ is the angle between the water flow and the ship.

[0089] Based on the load data, real-time navigation parameters, environmental parameters, and reward values, a CART algorithm is used for preliminary classification. Specifically, this data is used as input features. The CART algorithm recursively partitions the feature space, selects the optimal split point, generates a decision tree, divides the data into different categories, and outputs a classification feature vector containing the ship's operating status value. This vector is composed of factors such as the speed change rate and load ratio, and represents the ship's operating characteristics. Based on the classification feature vector, a support vector machine algorithm is used for vector optimization. Specifically, the classification feature vector is mapped to a high-dimensional space, and the optimal hyperplane is found by maximizing the classification interval. The optimized hyperplane is used as a modified classification boundary condition to more accurately distinguish different operating states. A multi-dimensional cluster analysis is performed based on the modified classification boundary conditions, the classification feature vectors and the reward values. The specific operation is to combine the modified classification boundary conditions, use a clustering algorithm such as K-means clustering, take the classification feature vectors and the reward values ​​as input, group them in a multi-dimensional space, and generate a load state classification result including normal load, overload and optimal load states. The load state classification result including normal load, overload and optimal load states is a ship load state category obtained based on cluster analysis. Normal load indicates that the load and navigation parameters are within the standard range, overload indicates that the load exceeds the safety range, and optimal load indicates the highest operating efficiency, which is used to analyze the ship acceleration performance and turning radius change trend in subsequent steps and serves as basic data for dynamic behavior analysis.

[0090] In step S12, it is necessary to calculate acceleration characteristics and perform turning radius regression analysis based on the load state classification result, the real-time navigation parameters and the environmental parameters to obtain acceleration performance indicators and turning radius change trends under different load states.

[0091] In one implementation, based on the load state classification result, the real-time navigation parameters, and the environmental parameters, acceleration characteristics are calculated and turning radius regression analysis is performed to obtain acceleration performance indicators and turning radius change trends under different load states, including:

[0092] According to the load state classification result, the real-time navigation parameters and the environmental parameters, classification calculations are performed to obtain the ship mass distribution parameters and the hydrodynamic resistance coefficient; according to the real-time navigation parameters and the ship mass distribution parameters, the ship acceleration and the acceleration response time are calculated, and the obtained ship acceleration and acceleration response time are subjected to time series analysis to obtain acceleration performance indicators under different load states; according to the ship mass distribution parameters, the hydrodynamic resistance coefficient, the environmental parameters and the real-time navigation parameters, a regression analysis is performed to infer the mapping relationship between the load state and the turning radius change rate, and the turning radius change trend is obtained.

[0093] It should be noted that, according to the load state classification result, the real-time navigation parameters and the environmental parameters, classification calculations are performed to obtain the ship mass distribution parameters and the hydrodynamic resistance coefficient. The specific operation is to decompose the load data into mass distribution of different parts according to the load state classification result, such as normal load or overload, and calculate the hydrodynamic resistance coefficient through a preset fluid mechanics model in combination with real-time navigation parameters such as speed and environmental parameters such as water flow velocity. Based on the ship's geometric shape and fluid mechanics principles, the ship's hull parameters, such as length, width, draft and hull surface roughness, are first obtained according to the ship design drawings or actual measurement data. The influence of the ship's current load distribution on the draft is determined in combination with the load state classification result. The real-time navigation parameters provide the ship's speed, and the environmental parameters provide water flow velocity and water density, etc. Information is obtained and these parameters are input into the fluid mechanics model, which is based on the Reynolds-averaged Navier-Stokes equations. It takes into account the turbulent characteristics of the flow field around the ship, and solves the flow field distribution through numerical simulation methods such as the finite volume method. The friction resistance and pressure resistance on the ship surface are calculated. The friction resistance is mainly determined by the surface roughness of the hull and the water velocity, while the pressure resistance is related to the hull shape and water flow separation. The friction resistance and pressure resistance are added together to obtain the total hydrodynamic resistance. According to the definition of fluid mechanics, the total hydrodynamic resistance is divided by the square of the ship speed, the water density and the wet surface area of ​​the hull to obtain the hydrodynamic resistance coefficient. The hydrodynamic resistance coefficient reflects the resistance characteristics of the ship when sailing in the water, and is used for subsequent analysis of the acceleration performance of the ship and the trend of changes in the turning radius. The ship mass distribution parameter reflects the distribution of the load on the ship. According to the real-time navigation parameters and the ship mass distribution parameters, the ship acceleration and the acceleration response time are calculated, and the obtained ship acceleration and acceleration response time are subjected to time series analysis to obtain the acceleration performance index under different load states. The specific operation is to use the propulsion power and speed changes in the real-time navigation parameters, combined with the ship mass distribution parameters, to calculate the ship acceleration through Newton's second law, and the acceleration response time is obtained by recording the time required for the speed to increase from an initial value to a target value. Subsequently, the acceleration and acceleration response time are arranged in chronological order, and their change patterns over time are analyzed. The acceleration characteristics under different load states are extracted to generate the acceleration performance index, which reflects the dynamic response capability of the ship under different load states. According to the ship mass distribution parameters, the hydrodynamic resistance coefficient, the environmental parameters and the real-time navigation parameters, a regression analysis is performed to infer the mapping relationship between the load state and the turning radius change rate, and the turning radius change trend is obtained. The specific operation is to use the ship mass distribution parameters, the hydrodynamic resistance coefficient, environmental parameters such as wave height and real-time navigation parameters such as rudder angle as input, adopt a multivariate linear regression model, fit the relationship between these parameters and the turning radius change rate in the real-time navigation parameters through the least squares method, infer the influence of the load state on the turning radius change, and obtain the turning radius change trend.The turning radius variation trend characterizes the law of how the turning ability of a ship changes with conditions under different load states. It reflects the mapping relationship between the turning radius change rate and the load state. It is used in subsequent steps to establish a mathematical relationship model between the load state and the trajectory arc characteristics, which serves as an important basis for analyzing the dynamic behavior of the ship.

[0094] In step S13, it is necessary to analyze the dynamic behavior of the ship according to the load state classification result and the turning radius change trend, and establish a mathematical relationship model between the load state and the trajectory arc characteristics.

[0095] In one implementation, the dynamic behavior of the ship is analyzed based on the load state classification result and the turning radius change trend, and a mathematical relationship model between the load state and the trajectory arc characteristics is established, including:

[0096] According to the load state classification results and the turning radius variation trend, a time series analysis is performed to obtain dynamic correlation characteristics; according to the dynamic correlation characteristics, the load state classification results and the turning radius variation trend, a random forest algorithm is used to perform correlation analysis to obtain a core parameter combination that affects the trajectory arc characteristics; the core parameter combination includes load state, track curvature and steering angle rate; according to the core parameter combination, a multivariate regression equation is constructed, and the least squares method is used to fit the nonlinear relationship between load state and track curvature and steering angle rate to obtain a load-track regression model; according to the load-track regression model, the load state classification results and the turning radius variation trend, combined with the standard trajectory arc parameters in a preset ship maneuvering feature library, the load-track regression model is constrainedly optimized to obtain a mathematical relationship model between load state and trajectory arc characteristics.

[0097] It should be noted that a time series analysis is performed based on the load state classification results and the turning radius variation trend to obtain dynamic correlation features. Specifically, the load state classification results (e.g., normal load or overload) and the turning radius variation trend are arranged in chronological order. A sliding window is used to slide a fixed time window across the time series. The correlation features between the load state and the turning radius variation trend within each window are calculated. The time-varying correlation patterns are extracted to generate dynamic correlation features that reflect the dynamic relationship between the load state and the turning radius variation. A random forest algorithm is used to perform correlation analysis based on the dynamic correlation features, the load state classification results, and the turning radius variation trend to obtain the core parameter combination that affects the trajectory arc characteristics. Specifically, the dynamic correlation features, the load state classification results, and the turning radius variation trend are used as input. The random forest algorithm constructs multiple decision trees, each trained based on randomly sampled features and data, calculates the importance of each feature on trajectory arc characteristics such as curvature and steering angle, and selects the most important core parameter combination, including load state, track curvature, and steering angle rate. According to the core parameter combination, a multiple regression equation is constructed, and the least squares method is used to fit the nonlinear relationship between the load state and the track curvature and the steering angular rate to obtain the load-track regression model. The specific operation is to use the load state, track curvature and steering angular rate as core variables to construct a multiple regression equation. It is assumed that the track curvature and steering angular rate have a nonlinear relationship with the load state. The equation is constructed in polynomial or exponential form, and the least squares method is used to optimize the parameters to minimize the sum of squares of the errors between the predicted value and the actual value. The load-track regression model is generated to describe the preliminary influence of the load state on the trajectory arc characteristics. Based on the load-track regression model, the load state classification results, and the turning radius variation trend, combined with the standard trajectory arc parameters in the preset ship maneuvering feature library, the load-track regression model is constrained and optimized to obtain a mathematical relationship model between the load state and the trajectory arc characteristics. The specific operation is to compare the prediction results of the load-track regression model with the standard trajectory arc parameters in the ship maneuvering feature library, such as typical curvature values ​​and steering angle rate ranges, set constraints such as the curvature must not exceed the physical limit, and combine the load state classification results and the turning radius variation trend to adjust the model parameters through the gradient descent method so that the model meets the constraints and generates a mathematical relationship model between the load state and the trajectory arc characteristics. The mathematical relationship model between the load state and the trajectory arc characteristics is a mathematical expression that describes the precise relationship between the load state and the track curvature and steering angle rate. It is used in the subsequent steps to analyze the dynamic heading adjustment characteristics of the dense navigation area and serves as the basis for generating the dynamic correlation matrix.

[0098] In step S14, it is necessary to analyze the dynamic heading adjustment characteristics of the dense navigation area based on the mathematical relationship model between the load state and the trajectory arc characteristics, the acceleration performance index, the real-time navigation parameters and the environmental parameters, perform correlation modeling, and generate a dynamic correlation matrix between the load state and the trajectory characteristics.

[0099] In one implementation, based on the mathematical relationship model between the load state and trajectory arc characteristics, the acceleration performance index, the real-time navigation parameters, and the environmental parameters, the dynamic heading adjustment characteristics of the dense navigation area are analyzed, and correlation modeling is performed to generate a dynamic correlation matrix between the load state and trajectory characteristics, including:

[0100] According to the mathematical relationship model between the load state and the trajectory arc characteristics, the real-time navigation parameters and the environmental parameters, a multivariate collaborative analysis is performed to obtain the heading adjustment response speed and the heading adjustment amplitude; according to the acceleration performance index and the real-time navigation parameters, the dynamic maneuvering margin of the ship in the dense navigation area is calculated, and the heading adjustment urgency evaluation index is constructed in combination with the preset collision avoidance safety distance threshold; according to the mathematical relationship model between the load state and the trajectory arc characteristics, the heading adjustment response speed and the heading adjustment amplitude, a convolutional neural network is used to perform spatiotemporal correlation analysis to identify the implicit influence relationship of the load state on the heading adjustment mode; according to the implicit influence relationship, the acceleration performance index, the real-time navigation parameters and the environmental parameters, multivariate linear regression is used to perform multi-scale matching to establish a three-dimensional mapping relationship between the load state, acceleration performance and heading adjustment sensitivity; according to the three-dimensional mapping relationship and the heading adjustment urgency evaluation index, K-means clustering analysis is used to perform correlation modeling to obtain a dynamic correlation matrix between the load state and the trajectory characteristics.

[0101] It should be noted that, based on the mathematical relationship model of the load state and the trajectory arc characteristics, the real-time navigation parameters and the environmental parameters, a multivariate collaborative analysis is performed to obtain the heading adjustment response speed and the heading adjustment amplitude. The specific operation is to combine the track curvature and turning angle rate predicted by the mathematical relationship model with the real-time navigation parameters such as speed, rudder angle and environmental parameters such as wind speed and water flow velocity. Through multivariate statistical analysis, the heading adjustment response speed, that is, the time required for the ship to complete the heading adjustment from receiving the instruction to completing the heading adjustment, and the heading adjustment amplitude, that is, the angle of the heading change, are calculated to reflect the ship's heading adjustment capability in dense navigation areas. According to the acceleration performance index and the real-time navigation parameters, the dynamic maneuvering margin of the ship in the dense navigation area is calculated, and combined with the preset collision avoidance safety distance threshold, a heading adjustment urgency evaluation index is constructed. The specific operation is to use the acceleration and acceleration response time in the acceleration performance index, combined with real-time navigation parameters such as speed, to calculate the adjustable space and time margin of the ship in the current state, and obtain the dynamic maneuvering margin, and then compare it with the preset collision avoidance safety distance threshold. When the dynamic maneuvering margin is less than the preset collision avoidance safety distance threshold, an emergency adjustment analysis of the ship's heading is performed, and a heading adjustment urgency evaluation index is generated to reflect the urgency of the adjustment. Based on the mathematical relationship model between the load state and trajectory arc characteristics, the heading adjustment response speed, and the heading adjustment amplitude, a convolutional neural network is used to perform spatiotemporal correlation analysis to identify the implicit influence of the load state on the heading adjustment mode. Specifically, the output of the mathematical relationship model, the heading adjustment response speed, and the heading adjustment amplitude are used as input. The convolutional neural network extracts spatiotemporal features through the convolution layer. Combined with the pooling layer and the fully connected layer, it analyzes the deep impact of the load state on the heading adjustment mode, such as the adjustment frequency and amplitude, and identifies the implicit influence of the load state on the heading adjustment mode. Based on the implicit influence relationship, the acceleration performance index, the real-time navigation parameters, and the environmental parameters, multivariate linear regression is used for multi-scale matching to establish a three-dimensional mapping relationship between load state, acceleration performance, and heading adjustment sensitivity. Specifically, the multivariate linear regression is used to fit the data at different time and space scales using the least squares method to generate a three-dimensional mapping relationship between load state, acceleration performance, and heading adjustment sensitivity. According to the three-dimensional mapping relationship and the heading adjustment urgency evaluation index, K-means clustering analysis is used to perform association modeling to obtain a dynamic association matrix between load status and trajectory characteristics. The specific operation is to use the three-dimensional mapping relationship and the heading adjustment urgency evaluation index as input, and the K-means clustering algorithm is iteratively optimized to divide the data into multiple clusters. Each cluster reflects the association pattern between load status and trajectory characteristics such as curvature and heading change, thereby generating a dynamic association matrix.The dynamic correlation matrix between load states and trajectory characteristics is a matrix structure that reflects the dynamic correlation between trajectory characteristics under different load states. It is obtained through the above-mentioned K-means clustering analysis and is used to dynamically adjust the weight values ​​of trajectory similarity calculation in subsequent steps as the basic data for weight allocation.

[0102] In step S15 , it is necessary to dynamically adjust the weight value of the load state in the trajectory similarity calculation according to the dynamic association matrix, the real-time navigation parameters and the environmental parameters to obtain a weight distribution result.

[0103] In one implementation, dynamically adjusting the weight value of the load state in trajectory similarity calculation according to the dynamic correlation matrix, the real-time navigation parameters, and the environmental parameters to obtain a weight distribution result includes:

[0104] According to the dynamic association matrix, adaptive weight allocation is performed to construct a weight allocation feature space, including load state weight, heading adjustment priority and trajectory deviation tolerance; according to the load state weight, the real-time navigation parameters and the environmental parameters, a Bayesian optimization algorithm is used to establish a weight allocation strategy, dynamically adjust the basic allocation ratio of the load state weight, and obtain a first load state weight; according to the heading adjustment priority, the real-time navigation parameters and the environmental parameters, combined with the safe operation threshold in the preset ship collision avoidance rule library, the heading adjustment priority is normalized to generate a dynamic constraint condition for weight adjustment; according to the trajectory deviation The difference tolerance and the first load state weight are learned in real time using a reinforcement learning algorithm. When the risk of intersection of adjacent ship trajectories is detected, the load state weight is increased in real time to obtain a second load state weight. Trajectory similarity calculation is performed based on the second load state weight, the dynamic constraint conditions, the real-time navigation parameters, and the environmental parameters. When the trajectory similarity is less than a preset trajectory similarity threshold, the load state weight is dynamically corrected in combination with the dynamic constraint conditions to obtain a weight allocation result. When the trajectory similarity is greater than the preset trajectory similarity threshold, the second load state weight is determined as the final load state weight to obtain a weight allocation result.

[0105] It should be noted that, according to the dynamic association matrix, adaptive weight allocation is performed to construct a weight allocation feature space, including load state weight, heading adjustment priority and trajectory deviation tolerance. The specific operation is to extract the correlation data between load state and trajectory characteristics from the dynamic association matrix, analyze the influence of load state on trajectory characteristics, preliminarily allocate load state weights, and construct a multidimensional feature space, including load state weights, heading adjustment priority and trajectory deviation tolerance, reflecting the role of different factors in weight allocation. According to the load state weight, the real-time navigation parameters and the environmental parameters, a Bayesian optimization algorithm is used to establish a weight allocation strategy, dynamically adjust the basic allocation ratio of the load state weight, and obtain the first load state weight. The specific operation is to combine the preliminary load state weight with real-time navigation parameters such as speed, rudder angle and environmental parameters such as wind speed and water flow speed. The Bayesian optimization algorithm predicts the performance of weight allocation by constructing a Gaussian process model, iteratively optimizes the weight ratio, and generates a first load state weight, reflecting the importance of the load state under the current navigation conditions. Based on the course adjustment priority, the real-time navigation parameters, and the environmental parameters, and in combination with the safety operation threshold in the preset ship collision avoidance rule library, the course adjustment priority is normalized to generate a dynamic constraint for weight adjustment. Specifically, the course adjustment priority is calculated based on the real-time navigation parameters and environmental parameters. In combination with the safety operation threshold in the preset ship collision avoidance rule library, such as the minimum safety distance, the course adjustment priority is mapped to a range of 0 to 1, and a normalized priority value is generated as a dynamic constraint for limiting the weight adjustment range. Based on the trajectory deviation tolerance and the first load state weight, a reinforcement learning algorithm is used to perform real-time learning. When the risk of adjacent ship trajectory intersection is detected, the load state weight is increased in real time to obtain a second load state weight. Specifically, the trajectory deviation tolerance and the first load state weight are used as input. The reinforcement learning algorithm uses the Q learning method to set the reward function to avoid trajectory intersection. When the real-time navigation data detects that the adjacent ship trajectory may intersect, the load state weight is increased to generate a second load state weight that reflects the adjustment of the weight to safety requirements.Trajectory similarity is calculated based on the second load state weight, the dynamic constraints, the real-time navigation parameters, and the environmental parameters. When the trajectory similarity is less than a preset trajectory similarity threshold, the load state weight is dynamically modified in combination with the dynamic constraints to obtain a weight allocation result. When the trajectory similarity is greater than the preset trajectory similarity threshold, the second load state weight is determined as the final load state weight to obtain a weight allocation result. Specifically, the second load state weight is weighted with the real-time navigation parameters and environmental parameters to calculate the similarity between the current trajectory and the target trajectory. If the similarity is less than the preset threshold, the weight is adjusted based on the dynamic constraints and recalculated until the conditions are met, generating a weight allocation result. If the similarity is greater than the threshold, the second load state weight is directly used as the weight allocation result. The weight allocation result is a final weight value reflecting the importance of the load state in the trajectory similarity calculation and is used in the weighted clustering analysis and similarity calculation in subsequent steps as the basis for trajectory similarity determination.

[0106] In step S16, cluster analysis is performed based on the weight distribution result and the real-time navigation parameters, and the similarity of similar load state trajectories is calculated in combination with the preset path constraint conditions to obtain the final similarity judgment result.

[0107] In one implementation, cluster analysis is performed based on the weight distribution result and the real-time navigation parameters, and the similarity of trajectories with similar load states is calculated in combination with preset path constraints to obtain a final similarity judgment result, including:

[0108] Based on the weight distribution result and the real-time navigation parameters, the navigation trajectory characteristics are extracted, the trajectory normalization processing is performed, and a trajectory feature vector data set is generated; based on the trajectory feature vector data set, a density clustering algorithm is used to perform spatial distribution analysis, and classification is performed according to the load state weight ratio to obtain trajectory clusters containing high similarity, medium similarity, and low similarity; based on the trajectory clusters, dynamic time warping processing is performed, and the dynamic bending distance between trajectory sequences is calculated in combination with the minimum turning radius threshold and the safe speed limit in the preset path constraint conditions; based on the dynamic bending distance and the weight distribution result, a risk compensation analysis is performed, a multidimensional similarity evaluation index is constructed, and the trajectories in the same cluster are sorted by similarity to obtain a similarity sorting result; based on the similarity sorting result and the real-time navigation parameters, spatiotemporal alignment is performed to obtain a final similarity judgment result including load state matching, path consistency coefficient, and risk warning level.

[0109] It should be noted that, based on the weight distribution results and the real-time navigation parameters, the navigation trajectory features are extracted, the trajectory is normalized, and a trajectory feature vector dataset is generated. The specific operation is to extract trajectory features such as the course change rate and speed fluctuation from the real-time navigation parameters such as speed, heading, and position coordinates, weight the features in combination with the load state weight in the weight distribution results, and map these features to a unified range through normalization to generate a multi-dimensional trajectory feature vector dataset that reflects the comprehensive characteristics of the ship's trajectory. Based on the trajectory feature vector dataset, a density clustering algorithm is used to perform spatial distribution analysis, and classification is performed based on the load state weight ratio to obtain trajectory clusters containing high similarity, medium similarity, and low similarity. The specific operation is to input the trajectory feature vector dataset into a density clustering algorithm such as DBSCAN, calculate the Euclidean distance between vectors, group the data points based on density accessibility, and classify the groups based on the load state weight ratio to generate high, medium, and low similarity trajectory clusters that reflect the similarity distribution between trajectories. Based on the trajectory cluster, dynamic time warping is performed, and the dynamic curvature distance between trajectory sequences is calculated by combining the minimum turning radius threshold and the safe speed limit in the preset path constraint conditions. Specifically, for the trajectory sequence in each trajectory cluster, a dynamic time warping algorithm is used to align the time series of the two trajectories through a dynamic programming method, calculate the optimal matching path, combine the minimum turning radius threshold and the safe speed limit to constrain the matching path, ensure that the trajectory meets the physical constraints, and generate a dynamic curvature distance to reflect the similarity of the trajectory in the time dimension. Based on the dynamic curvature distance and the weight distribution result, a risk compensation analysis is performed, a multidimensional similarity evaluation index is constructed, and the trajectories in the same cluster are sorted by similarity to obtain a similarity sorting result. Specifically, the dynamic curvature distance is used, combined with the load state weight in the weight distribution result, to calculate the weighted distance, and at the same time analyze the trajectory intersection risk, set the risk compensation coefficient to adjust the distance, construct a multidimensional similarity evaluation index such as distance and risk weighted value, sort the trajectories in the same cluster by similarity, and generate a similarity sorting result. According to the similarity sorting result and the real-time navigation parameters, time and space alignment is performed to obtain a final similarity judgment result including load state matching, path consistency coefficient and risk warning level. The specific operation is to combine the similarity sorting result with real-time navigation parameters such as position and speed, align the time and space dimensions, calculate the load state matching, that is, the similarity of the load states of the two trajectories, the path consistency coefficient, that is, the degree of fit of the trajectory paths, and the risk warning level based on the trajectory intersection risk assessment to generate the final similarity judgment result.The final similarity judgment result, which includes load state matching, path consistency coefficient and risk warning level, is a conclusion of comprehensive evaluation of the similarity of the two trajectories. The load state matching reflects the similarity of the load states, the path consistency coefficient reflects the path consistency, and the risk warning level reflects the potential risk. It is used for navigation optimization, safety warning and trajectory comparison analysis, and serves as the basis for ship navigation decision-making.

[0110] To facilitate understanding of the present invention, some preferred embodiments of the present invention are further described below.

[0111] The following describes the working process of the present invention using a relatively common scenario as an example.

[0112] At the intersection of the main channels of a certain international hub port, a fully loaded container ship is sailing towards the berth at an economic speed. At this time, there are three incoming cargo ships and two outgoing oil tankers in the channel at the same time. The sea condition is affected by the tide and shows an increasing flow rate. Traditional navigation systems usually use a fixed safety distance threshold for trajectory prediction in such dense navigation areas, but it is difficult to cope with the changes in the maneuverability of the ship caused by the difference in load. The method of the present invention collects the load data, speed and heading parameters and surrounding flow speed and wind direction information of the container ship in real time through ship-borne sensors. First, based on a preset reward mechanism, an economic evaluation of the current operating status of the ship is performed, and a comprehensive load value is generated by combining the load distribution, draft depth and fuel consumption rate. Then, a decision tree algorithm is used to divide the real-time navigation status into high-load operation categories, and after optimizing the classification boundaries through a support vector machine, the ship is finally clustered and identified as being in a fully loaded optimal load state.

[0113] Afterwards, the current ship mass distribution parameters and hydrodynamic resistance coefficient were retrieved based on the classification results. Combined with the propeller output power and real-time flow rate data, it was calculated that the acceleration response time of the ship under full load was 40% longer than that under standard working conditions. At the same time, based on the rudder efficiency model, the turning radius was deduced to be 35% higher than that under no-load conditions. These dynamic parameters were subjected to multivariate regression analysis. The sliding window was used to extract the steering angle rate and curvature change characteristics in the nearly ten-minute heading adjustment sequence. The random forest algorithm was used to screen out the three core influencing factors of load offset, heel angle fluctuation, and propeller response delay, and a nonlinear mapping relationship between load state and trajectory arc was constructed. At this time, a departing oil tanker in front began to deviate from the scheduled route due to a sudden mechanical failure. After the system sensed the risk of track crossing through AIS data, it immediately activated the dynamic heading adjustment analysis module in the dense navigation area.

[0114] First, based on a mathematical relationship model, the curvature of the standard avoidance trajectory of the container ship under the current full load state is predicted. Combined with the real-time cruise ship drift speed and azimuth data, a convolutional neural network is used to analyze the spatiotemporal correlation of the two ships' course adjustment modes, identifying the key nodes that require the avoidance action to be initiated 15 seconds in advance under the current load state. At the same time, the dynamic maneuvering margin of the ship under the influence of tidal flow is calculated based on the acceleration performance index. Combined with the safety distance threshold set by the International Collision Prevention Regulations, an assessment result with a course adjustment urgency level of level 2 is generated. At this time, the system automatically constructs a three-dimensional mapping relationship, associates and matches the load state's influence coefficient on rudder efficiency, the propulsion system response delay, and the course correction sensitivity, and generates a dynamic correlation matrix containing load weight, trajectory deviation tolerance, and avoidance priority.

[0115] During the weight allocation phase, the basic proportion of load state weights is dynamically adjusted based on a Bayesian optimization algorithm, initially setting the fully loaded state's trajectory feature weight at 65%. When the tanker's drift speed is detected to exceed a preset threshold, the reinforcement learning module immediately intervenes, increasing the load weight to 78% based on the dynamic maneuvering margin calculated in real time. Simultaneously, a weight compensation mechanism is triggered based on the response speed and amplitude of the heading adjustment to correct trajectory deviations caused by sudden tidal changes. Monte Carlo simulations are then used to verify the alignment of the adjusted weight allocation strategy with the collision avoidance safety objectives. Once compliance is confirmed, the updated weight parameters are used for trajectory similarity calculations.

[0116] During the trajectory matching phase, the system first normalizes the avoidance trajectories of similar fully loaded vessels under similar sea conditions over the past three months to generate a trajectory vector set containing load characteristic weights. A density clustering algorithm is then used to partition the trajectory library into three clusters with high, medium, and low similarity. The high-similarity cluster contains 21 avoidance trajectories that match the current tidal intensity and load distribution. The dynamic time warping algorithm is used to process the current vessel's real-time trajectory sequence. Combining minimum turning radius constraints with safe speed limits, the dynamic bending distance from the representative trajectory of each cluster is calculated. Matching is prioritized within the high-similarity cluster, identifying three historical trajectories with over 85% agreement with the current course adjustment pattern. The similarity scores are then adjusted based on a risk compensation factor. The final output shows the optimal matching trajectory with a load state match of 91%, a path consistency coefficient of 0.87, and a yellow risk warning level. A simultaneous avoidance recommendation is generated: turn 10 degrees to starboard and reduce speed to 12 knots.

[0117] During this process, the dynamic correlation matrix is ​​updated every two seconds. Upon detecting that the tanker has regained power, the load status weight is immediately reduced to the baseline level, and the urgency flag for course adjustment is removed. Throughout the avoidance maneuver, this invention dynamically adjusts the weight allocation strategy by correlating load status with trajectory characteristics in real time. This method successfully controls the trajectory similarity calculation error to within 3.2%, improving accuracy by 41% compared to traditional methods and effectively avoiding the risk of misjudgment caused by ignoring load status.

[0118] In summary, the present invention discloses a method for judging the similarity of ship trajectories based on multi-dimensionality and dynamic weights, comprising obtaining load data, real-time navigation parameters and environmental parameters in ship navigation data, performing analysis based on a reward mechanism, and generating a load state classification result; calculating acceleration characteristics and performing a turning radius regression analysis based on the load state classification result, the real-time navigation parameters and the environmental parameters, and obtaining acceleration performance indicators and turning radius change trends under different load states; analyzing the dynamic behavior of the ship based on the load state classification result and the turning radius change trend, and establishing a mathematical relationship model between the load state and the trajectory arc characteristics. ; According to the mathematical relationship model between the load state and the trajectory arc characteristics, the acceleration performance index, the real-time navigation parameters and the environmental parameters, the dynamic heading adjustment characteristics of the dense navigation area are analyzed, and correlation modeling is performed to generate a dynamic correlation matrix of the load state and trajectory characteristics; according to the dynamic correlation matrix, the real-time navigation parameters and the environmental parameters, the weight value of the load state in the trajectory similarity calculation is dynamically adjusted to obtain a weight distribution result; according to the weight distribution result and the real-time navigation parameters, a cluster analysis is performed, and combined with the preset path constraint conditions, the similarity of trajectories with similar load states is calculated to obtain a final similarity judgment result.

[0119] The present invention analyzes the dynamic coupling relationship between ship load, speed and environmental parameters in real time by constructing a load state classification model based on a reward mechanism, and then uses a dual model of acceleration and turning radius to collaboratively calculate and quantify the differences in ship maneuverability under different load states. Based on the multivariate regression analysis method, a nonlinear mapping relationship between load state and trajectory arc curvature is established. At the same time, a multidimensional dynamic correlation matrix is ​​generated by combining the spatiotemporal correlation modeling of the dynamic heading adjustment characteristics of dense navigation areas. In this process, an adaptive weight allocation algorithm driven by the dual engines of Bayesian optimization and reinforcement learning is used to dynamically adjust the load state weight ratio according to real-time speed fluctuations, avoidance requirements and environmental disturbances. Finally, under the path constraint condition, a dynamic time warping algorithm is used to fuse multidimensional features to realize trajectory similarity calculation, and then spatiotemporally align with real-time navigation parameters to obtain the final similarity judgment result including load state matching degree, path consistency coefficient and risk warning level. This breaks through the limitations of the static modeling and fixed weight allocation mechanism of the traditional method, realizes the dynamic association of ship load state with trajectory characteristics, and realizes adaptive weight adjustment in the similarity judgment process.

[0120] Based on the same inventive concept as the method provided in the embodiment of this application, this embodiment also provides a ship trajectory similarity judgment device based on multi-dimensional and dynamic weights. For any unclear points in the embodiment of the device, please refer to the content in the above method embodiment. Figure 2 The second embodiment of the present invention provides a device for determining the similarity of ship trajectories based on multi-dimensionality and dynamic weights, comprising:

[0121] The data acquisition module is used to obtain load data, real-time navigation parameters and environmental parameters from the ship's navigation data, analyze them based on the reward mechanism, and generate load status classification results;

[0122] a speed-path analysis module, configured to calculate acceleration characteristics and perform turning radius regression analysis based on the load state classification result, the real-time navigation parameters, and the environmental parameters, to obtain acceleration performance indicators and turning radius change trends under different load states;

[0123] A regression modeling module is used to analyze the dynamic behavior of the ship based on the load state classification results and the turning radius change trend, and to establish a mathematical relationship model between the load state and the trajectory arc characteristics;

[0124] a dynamic matrix module for analyzing the dynamic course adjustment characteristics of the dense navigation area based on the mathematical relationship model between the load state and the trajectory arc characteristics, the acceleration performance index, the real-time navigation parameters, and the environmental parameters, performing correlation modeling, and generating a dynamic correlation matrix between the load state and the trajectory characteristics;

[0125] a weight distribution module, configured to dynamically adjust the weight value of the load state in the trajectory similarity calculation according to the dynamic association matrix, the real-time navigation parameters, and the environmental parameters to obtain a weight distribution result;

[0126] The result output module is used to perform cluster analysis based on the weight distribution result and the real-time navigation parameters, and calculate the similarity of similar load state trajectories in combination with preset path constraints to obtain a final similarity judgment result.

[0127] It should be noted that the device for determining the similarity of ship trajectories based on multi-dimensionality and dynamic weights provided in an embodiment of the present invention is used to execute all the process steps of the method for determining the similarity of ship trajectories based on multi-dimensionality and dynamic weights in the above embodiment. The working principles and beneficial effects of the two correspond one to one, and therefore will not be described in detail.

[0128] An embodiment of the present invention further provides an electronic device. The electronic device includes: a processor, a memory, and a computer program stored in the memory and executable on the processor, such as a program for determining the similarity of ship trajectories based on multiple dimensions and dynamic weights. When the processor executes the computer program, the steps in each of the above-mentioned embodiments of the method for determining the similarity of ship trajectories based on multiple dimensions and dynamic weights are implemented, such as Figure 1 Alternatively, when the processor executes the computer program, the functions of the modules / units in the above-mentioned device embodiments are realized, such as the weight allocation module.

[0129] Exemplarily, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to implement the present invention. The one or more modules / units may be a series of computer program instruction segments capable of implementing specific functions, and the instruction segments are used to describe the execution process of the computer program in the electronic device.

[0130] The electronic device may be a computing device such as a desktop computer, notebook, PDA, or smart tablet. The electronic device may include, but is not limited to, a processor and memory. Those skilled in the art will appreciate that the aforementioned components are merely examples of electronic devices and do not constitute a limitation of the electronic device. The electronic device may include more or fewer components than those described above, or a combination of certain components, or different components. For example, the electronic device may also include input / output devices, network access devices, buses, and the like.

[0131] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the electronic device, connecting various parts of the entire electronic device using various interfaces and lines.

[0132] The memory can be used to store the computer programs and / or modules, and the processor realizes various functions of the electronic device by running or executing the computer programs and / or modules stored in the memory and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, at least one application required for a function (such as a sound playback function, an image playback function, etc.); the data storage area can store data created based on the use of the mobile phone (such as audio data, a phone book, etc.). In addition, the memory can include a high-speed random access memory and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage device.

[0133] Wherein, if the module / unit integrated in the electronic device is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the process in the above-mentioned embodiment method, and can also be completed by a computer program to instruct the relevant hardware. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by the processor, it can implement the steps of each of the above-mentioned method embodiments. Wherein, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.

[0134] It should be noted that the device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed across multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present embodiment. In addition, in the drawings of the device embodiments provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines. A person of ordinary skill in the art can understand and implement the present invention without inventive effort.

[0135] The specific embodiments described above further illustrate the objectives, technical solutions, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A method for judging the similarity of ship trajectories based on multi-dimensionality and dynamic weights, characterized in that: include: Obtain load data, real-time navigation parameters, and environmental parameters from the ship's navigation data, analyze them based on the reward mechanism, and generate load status classification results; Calculating acceleration characteristics and performing turning radius regression analysis based on the load state classification result, the real-time navigation parameters, and the environmental parameters to obtain acceleration performance indicators and turning radius change trends under different load states; Analyzing the dynamic behavior of the ship based on the load state classification results and the turning radius change trend, and establishing a mathematical relationship model between the load state and the trajectory arc characteristics; According to the mathematical relationship model between the load state and the trajectory arc characteristics, the acceleration performance index, the real-time navigation parameters and the environmental parameters, the dynamic course adjustment characteristics of the dense navigation area are analyzed, and correlation modeling is performed to generate a dynamic correlation matrix between the load state and the trajectory characteristics; According to the dynamic correlation matrix, the real-time navigation parameters and the environmental parameters, the trajectory similarity is dynamically adjusted to calculate the weight value of the load state and obtain a weight distribution result; A cluster analysis is performed based on the weight distribution result and the real-time navigation parameters. The similarity of trajectories with similar load states is calculated in combination with preset path constraints to obtain a final similarity judgment result.

2. The method for determining ship trajectory similarity based on multi-dimensionality and dynamic weights according to claim 1 is characterized in that: The method of obtaining load data, real-time navigation parameters and environmental parameters from the ship navigation data, analyzing the load data based on the reward mechanism, and generating a load status classification result includes: Obtain load data, real-time navigation parameters and environmental parameters from ship navigation data; According to the load data, the real-time navigation parameters and the environmental parameters, a preset reward mechanism is used to calculate a reward value; wherein the size of the reward value reflects the quality of the ship's operating status; Performing preliminary classification using a CART algorithm based on the load data, the real-time navigation parameters, the environmental parameters, and the reward value to obtain a classification feature vector containing a ship operation status value; According to the classification feature vector, a support vector machine algorithm is used to perform vector optimization to obtain a modified classification boundary condition; A multi-dimensional cluster analysis is performed based on the modified classification boundary conditions, the classification feature vectors, and the reward value to obtain a load state classification result including normal load, overload, and optimal load states.

3. The method for determining ship trajectory similarity based on multi-dimensionality and dynamic weights according to claim 1, characterized in that: The calculation of acceleration characteristics and the performance of turning radius regression analysis based on the load state classification result, the real-time navigation parameters, and the environmental parameters to obtain acceleration performance indicators and turning radius change trends under different load states includes: Performing classification calculations based on the load state classification results, the real-time navigation parameters, and the environmental parameters to obtain ship mass distribution parameters and hydrodynamic resistance coefficients; Calculating the ship acceleration and acceleration response time based on the real-time navigation parameters and the ship mass distribution parameters, performing time series analysis on the obtained ship acceleration and acceleration response time, and obtaining acceleration performance indicators under different load conditions; A regression analysis is performed based on the ship mass distribution parameters, the hydrodynamic resistance coefficient, the environmental parameters and the real-time navigation parameters to infer a mapping relationship between the load state and the turning radius change rate, thereby obtaining a turning radius change trend.

4. The method for determining ship trajectory similarity based on multi-dimensionality and dynamic weights according to claim 1, characterized in that: The analyzing the dynamic behavior of the ship based on the load state classification result and the turning radius change trend, and establishing a mathematical relationship model between the load state and the trajectory arc characteristics, includes: Performing time series analysis based on the load state classification results and the turning radius change trend to obtain dynamic correlation characteristics; A random forest algorithm is used to perform correlation analysis based on the dynamic correlation characteristics, the load state classification results, and the turning radius change trend to obtain a core parameter combination that affects the trajectory arc characteristics; The core parameter combination includes load state, track curvature and steering angle rate; According to the core parameter combination, a multivariate regression equation is constructed, and the nonlinear relationship between the load state and the track curvature and the steering angle rate is fitted using the least squares method to obtain a load-track regression model; According to the load-track regression model, the load state classification result and the turning radius change trend, combined with the standard trajectory arc parameters in a preset ship maneuvering feature library, the load-track regression model is constrainedly optimized to obtain a mathematical relationship model between load state and trajectory arc characteristics.

5. The method for determining ship trajectory similarity based on multi-dimensionality and dynamic weights according to claim 1, characterized in that: The method of analyzing the dynamic course adjustment characteristics of a dense navigation area based on the mathematical relationship model between the load state and the trajectory arc characteristics, the acceleration performance index, the real-time navigation parameters, and the environmental parameters, performing correlation modeling, and generating a dynamic correlation matrix between the load state and the trajectory characteristics includes: Performing a multivariate collaborative analysis based on the mathematical relationship model between the load state and the trajectory arc characteristics, the real-time navigation parameters, and the environmental parameters to obtain a heading adjustment response speed and a heading adjustment amplitude; Calculate the dynamic maneuvering margin of the ship in a dense navigation area based on the acceleration performance index and the real-time navigation parameters, and construct a course adjustment urgency assessment index based on a preset collision avoidance safety distance threshold; Based on the mathematical relationship model between the load state and the trajectory arc characteristics, the heading adjustment response speed and the heading adjustment amplitude, a convolutional neural network is used to perform spatiotemporal correlation analysis to identify the implicit influence of the load state on the heading adjustment mode; Based on the implicit influence relationship, the acceleration performance index, the real-time navigation parameters and the environmental parameters, multi-scale matching is performed using multiple linear regression to establish a three-dimensional mapping relationship between load state, acceleration performance and heading adjustment sensitivity; According to the three-dimensional mapping relationship and the heading adjustment urgency evaluation index, K-means cluster analysis is used to perform correlation modeling to obtain a dynamic correlation matrix between load status and trajectory characteristics.

6. The method for determining ship trajectory similarity based on multi-dimensionality and dynamic weights according to claim 1, characterized in that: The dynamically adjusting the weight value of the load state in the trajectory similarity calculation according to the dynamic association matrix, the real-time navigation parameters, and the environmental parameters to obtain a weight distribution result includes: Performing adaptive weight allocation according to the dynamic correlation matrix to construct a weight allocation feature space, including load state weight, heading adjustment priority, and trajectory deviation tolerance; According to the load state weight, the real-time navigation parameters and the environmental parameters, a Bayesian optimization algorithm is used to establish a weight distribution strategy, and a basic distribution ratio of the load state weight is dynamically adjusted to obtain a first load state weight; According to the course adjustment priority, the real-time navigation parameters and the environmental parameters, combined with the safe operation threshold in the preset ship collision avoidance rule library, the course adjustment priority is normalized to generate a dynamic constraint condition for weight adjustment; Based on the trajectory deviation tolerance and the first load state weight, a reinforcement learning algorithm is used to perform real-time learning. When a risk of crossing the trajectories of adjacent ships is detected, the load state weight is increased in real time to obtain a second load state weight. The trajectory similarity calculation is performed based on the second load state weight, the dynamic constraint condition, the real-time navigation parameter and the environmental parameter. When the trajectory similarity is less than a preset trajectory similarity threshold, the load state weight is dynamically corrected in combination with the dynamic constraint condition to obtain a weight distribution result. When the trajectory similarity is greater than the preset trajectory similarity threshold, the second load state weight is determined as the final load state weight to obtain a weight distribution result.

7. The method for determining ship trajectory similarity based on multi-dimensionality and dynamic weights according to claim 1, characterized in that: The cluster analysis is performed based on the weight distribution result and the real-time navigation parameters, and the dynamic time warping algorithm is used in combination with the path constraint condition to calculate the similarity of the trajectories with similar load states, and obtain the final similarity judgment result, including: Extracting navigation track features based on the weight distribution result and the real-time navigation parameters, performing track standardization processing, and generating a track feature vector data set; Based on the trajectory feature vector dataset, a density clustering algorithm is used to perform spatial distribution analysis, and classification is performed according to the load state weight ratio to obtain trajectory clusters including high similarity, medium similarity, and low similarity; Performing dynamic time warping on the trajectory clusters, and combining the minimum turning radius threshold and the safe speed limit in the preset path constraints to calculate the dynamic bending distance between the trajectory sequences; Perform risk compensation analysis based on the dynamic bending distance and the weight distribution result, construct a multi-dimensional similarity evaluation index, and perform similarity sorting on the trajectories in the same cluster to obtain a similarity sorting result; According to the similarity sorting result and the real-time navigation parameters, time and space alignment is performed to obtain a final similarity judgment result including load state matching degree, path consistency coefficient and risk warning level.

8. A ship trajectory similarity judgment system based on multi-dimensional and dynamic weights, characterized by: include: The data acquisition module is used to obtain load data, real-time navigation parameters and environmental parameters from the ship's navigation data, analyze them based on the reward mechanism, and generate load status classification results; a speed-path analysis module, configured to calculate acceleration characteristics and perform turning radius regression analysis based on the load state classification result, the real-time navigation parameters, and the environmental parameters, to obtain acceleration performance indicators and turning radius change trends under different load states; A regression modeling module is used to analyze the dynamic behavior of the ship based on the load state classification results and the turning radius change trend, and to establish a mathematical relationship model between the load state and the trajectory arc characteristics; a dynamic matrix module for analyzing the dynamic course adjustment characteristics of the dense navigation area based on the mathematical relationship model between the load state and the trajectory arc characteristics, the acceleration performance index, the real-time navigation parameters, and the environmental parameters, performing correlation modeling, and generating a dynamic correlation matrix between the load state and the trajectory characteristics; a weight distribution module, configured to dynamically adjust the weight value of the load state in the trajectory similarity calculation according to the dynamic association matrix, the real-time navigation parameters, and the environmental parameters to obtain a weight distribution result; The result output module is used to perform cluster analysis based on the weight distribution result and the real-time navigation parameters, and calculate the similarity of similar load state trajectories in combination with preset path constraints to obtain a final similarity judgment result.

9. An electronic device, characterized in that: The method comprises a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, the method for determining the similarity of ship trajectories based on multi-dimensionality and dynamic weights according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the ship trajectory similarity judgment method based on multi-dimensionality and dynamic weights according to any one of claims 1 to 7.

Images