A ship multi-source data interaction processing method and system based on a hub management platform

Abstract

This invention relates to the technical field of ship control, and in particular to a method and system for multi-source data interaction processing on ships based on a hub management platform. The method includes: achieving spatiotemporal synchronization of lidar, visual cameras, and millimeter-wave radar through timestamp alignment and coordinate mapping to construct a dynamic obstacle grid map; fusing multi-source data based on a dynamic Bayesian network, and dynamically adjusting sensor confidence weights by combining ambient light intensity and obstacle surface material; generating incremental obstacle avoidance strategies using a reinforcement learning model, and triggering graded response commands based on risk assessment levels. This invention solves the fusion error problem caused by the spatiotemporal asynchrony of multi-source data in ship navigation, and improves the real-time obstacle avoidance success rate of ships.

Description

Technical Field

[0001] This invention relates to the technical field of ship data processing, and in particular to a method and system for multi-source data interaction processing of ships based on a Hub management platform. Background Technology

[0002] With the increasing intelligence and informatization of ships, and the widespread application of electronic navigation equipment (such as ECDIS, AIS, and radar) and engine room monitoring systems, the multi-source and heterogeneous nature of navigation data is becoming increasingly significant. Existing ship management systems mostly adopt independent protocols and closed architectures, resulting in the following technical shortcomings:

[0003] 1. The problem of data silos is serious: Traditional solutions rely on device-specific protocols (such as NMEA 0183 one-way transmission) and lack standardized interfaces (such as RESTful API, OPC UA), which makes it difficult for data interaction between systems such as chart plotters, AIS, and radar, resulting in scattered navigation situation information and making it impossible to achieve cross-device collaborative analysis and operation.

[0004] 2. Low efficiency of heterogeneous data fusion: The formats of multi-source data (navigation data, environmental data, equipment status data) are very different (such as binary stream, JSON, XML). There is a lack of middleware layer for unified parsing and semantic alignment of data, making it difficult to build a centralized data warehouse and restricting real-time risk modeling and global decision-making.

[0005] 3. Existing systems rely on single-dimensional data analysis (such as AIS trajectory prediction) and fail to achieve deep correlation between navigation data (position, speed), environmental data (weather, ocean currents), and ship status data (engine room temperature, cargo hold humidity), thus failing to generate real-time navigation risk maps, resulting in delayed collision warnings or misjudgments of equipment malfunctions.

[0006] Therefore, in order to address the above-mentioned shortcomings, there is an urgent need for a comprehensive ship management platform that supports multi-protocol compatibility, deep data aggregation, and intelligent collaboration, in order to eliminate information silos, improve the accuracy of anomaly detection, and reduce human risk through data-driven operational suggestions. Summary of the Invention

[0007] To address the fusion error problem caused by the spatiotemporal asynchrony of multi-source data during ship navigation and to improve the real-time obstacle avoidance success rate of ships, this application provides a method and system for multi-source data interaction processing of ships based on a Hub management platform.

[0008] The above-mentioned objective of this application is achieved through the following technical solution:

[0009] A method for multi-source data interaction processing of ships based on a Hub management platform, the method comprising the following steps:

[0010] The raw sensor data from the lidar, vision camera and millimeter-wave radar on board the ship are acquired, and the raw sensor data is preprocessed to generate an obstacle grid map with unified spatiotemporal reference.

[0011] Based on a dynamic Bayesian network, multi-source data fusion is performed on the obstacle grid map to obtain dynamic motion trajectory prediction parameters for the obstacles.

[0012] The system acquires real-time ambient light intensity and obstacle surface material recognition results, and dynamically assigns confidence weights to each sensor based on these results.

[0013] Based on a preset reinforcement learning model, an incremental obstacle avoidance strategy set is generated according to the dynamic motion trajectory prediction parameters of the obstacle and the confidence weight of each sensor.

[0014] Based on the risk assessment level of the incremental obstacle avoidance strategy set, the corresponding navigation control command is triggered to complete the ship's autonomous obstacle avoidance.

[0015] By adopting the above technical solution, during navigation, the ship simultaneously collects timestamps, spatial coordinates, and signal strength data through lidar, visual cameras, and millimeter-wave radar, ensuring the integrity of the original input of multi-source data. The solution performs lidar point cloud noise reduction, visual image semantic segmentation, and millimeter-wave radar signal analysis to extract dynamic obstacle features. A dynamic Bayesian network is used to fuse multi-source data and output obstacle trajectory prediction parameters, such as position, velocity, and acceleration. Based on ambient light intensity and material identification, the sensor confidence level is dynamically adjusted. Material identification, such as the reflectivity of glass surfaces, is introduced to correct the state transition matrix, improving the detection success rate of transparent obstacles. Then, a priority obstacle avoidance strategy set is generated by combining trajectory prediction and weight vectors through a reinforcement learning model. Based on the risk assessment level of the obstacle avoidance strategy, control commands are triggered to achieve obstacle avoidance control of the ship. The entire technical solution, through multi-sensor collaboration and dynamic weight adjustment, significantly improves the obstacle avoidance success rate in complex environments, such as strong light and transparent obstacles, constructing a complete closed loop for ship obstacle avoidance technology.

[0016] In a preferred embodiment, this application can be further configured such that: the preprocessing of the raw sensor data to generate a spatiotemporally unified obstacle grid map specifically includes:

[0017] Gaussian filtering is applied to the point cloud data of the lidar to reduce noise and extract the three-dimensional boundary feature point set of the obstacle;

[0018] Perform semantic segmentation on images from a visual camera to identify obstacle category labels and surface texture feature vectors;

[0019] Fast Fourier transform is performed on the echo signal of millimeter-wave radar to analyze the radial velocity and distance information of obstacles;

[0020] The system acquires real-time pose data of the ship and generates a dynamic obstacle grid map in a unified coordinate system based on the real-time pose data, the three-dimensional boundary feature point set of obstacles, the obstacle category labels and surface texture feature vectors, the radial velocity and distance information of obstacles, and spatiotemporal interpolation compensation.

[0021] By employing the above technical solution, Gaussian filtering is applied to the point cloud data of the lidar to reduce noise and extract the three-dimensional boundary feature point set of obstacles. Semantic segmentation is performed on the images acquired by the visual camera to identify the category label and surface texture features of obstacles. Fast Fourier Transform is performed on the echo signal of the millimeter-wave radar to analyze the radial velocity and distance information of obstacles. Based on the real-time pose data of the ship, spatiotemporal interpolation compensation is used to fuse the data features output by various sensors, eliminating the spatiotemporal asynchrony and feature extraction incompatibility of multi-source sensor data, and generating a dynamic obstacle grid map in a unified coordinate system, which facilitates the subsequent generation of ship obstacle avoidance strategies and improves the accuracy of strategy generation.

[0022] In a preferred embodiment, this application can be further configured as follows: The acquisition of the ship's real-time pose data, based on the real-time pose data, the obstacle's three-dimensional boundary feature point set, obstacle category labels and surface texture feature vectors, and the obstacle's radial velocity and distance information, generates a dynamic obstacle grid map in a unified coordinate system through spatiotemporal interpolation compensation, specifically including:

[0023] A baseline time axis is established based on the LiDAR data cycle, and interpolation compensation is performed on the visual camera data to fill the time axis gaps.

[0024] Sliding window mean filtering is applied to millimeter-wave radar data to eliminate time asynchrony errors;

[0025] The compensated data is mapped onto a 3D grid map corresponding to the ship's real-time pose data to generate a dynamic obstacle grid map.

[0026] By adopting the above technical solution, taking the 10ms cycle of the lidar as the benchmark, aligning the data of other sensors, interpolating and compensating the data of the visual camera, solving the motion blur caused by insufficient visual frame rate, generating a 30fps continuous image stream, performing sliding window mean filtering on the millimeter-wave radar data, eliminating millimeter-wave radar signal jitter, suppressing speed change noise, converting multi-source data to the ship's body coordinate system, generating a dynamic obstacle grid map, and solving the problem of asynchronous fusion of multi-source sensor data.

[0027] In a preferred embodiment, this application can be further configured such that: the multi-source data fusion of the obstacle grid map based on a dynamic Bayesian network to obtain dynamic motion trajectory prediction parameters of the obstacles specifically includes:

[0028] Define an obstacle state transition matrix, which includes position, velocity, acceleration, and motion mode parameters;

[0029] Using the set of boundary feature points of the lidar as observation nodes, the probability distribution of the initial state of the obstacle is calculated.

[0030] Fuse the surface texture feature vectors of obstacles and update the material-related state transition probability weights;

[0031] Based on the radial velocity information, the trajectory prediction parameters of the dynamic Bayesian network are iteratively corrected.

[0032] By adopting the above technical solution and defining an obstacle state transition matrix, key parameters such as position, velocity, acceleration, and motion pattern are integrated, laying a solid foundation for subsequent trajectory prediction, improving prediction accuracy, and enhancing adaptability to complex obstacle motion patterns. The introduction of the state transition matrix enables a more comprehensive understanding of the obstacle's dynamic characteristics, thereby reducing uncertainty in prediction. Using the LiDAR boundary feature point set as observation nodes, the initial state probability distribution of the obstacle is calculated. The high-precision ranging capability of the LiDAR ensures the reliability of the initial state estimation. The high-resolution characteristics of the LiDAR data make the extraction of boundary feature points particularly accurate, further improving the accuracy of the initial state probability distribution and providing strong data support for subsequent dynamic Bayesian network inference. Finally, the surface texture feature vector of the obstacle is fused. The dynamic Bayesian network updates the state transition probability weights related to the material, enriching the descriptive dimensions of obstacle attributes. This not only reflects the physical characteristics of the obstacle but also indirectly reveals its possible motion characteristics, such as sliding and rolling. By fusing this information, the dynamic Bayesian network can more intelligently adjust the state transition probabilities, making trajectory prediction more realistic. The trajectory prediction parameters of the dynamic Bayesian network are iteratively corrected based on radial velocity information, realizing real-time tracking and dynamic adjustment of the obstacle's motion state. Radial velocity, as a key indicator that directly reflects the relative motion state of the obstacle, enables the prediction model to quickly respond to changes in the obstacle's velocity and adjust the predicted trajectory in a timely manner, thereby effectively reducing prediction errors and improving the real-time performance and accuracy of trajectory prediction. By comprehensively considering various attribute information of the obstacle, high-precision prediction of the dynamic motion trajectory of the obstacle is achieved.

[0033] In a preferred embodiment, this application can be further configured as follows: the real-time acquisition of ambient light intensity and obstacle surface material recognition results, and the dynamic allocation of confidence weights for each sensor based on the ambient light intensity and obstacle surface material recognition results, specifically includes:

[0034] The ambient light intensity is obtained based on the visual camera, and the weights of the visual sensor are dynamically adjusted based on the ambient light intensity.

[0035] Based on surface texture feature vector matching with a pre-set material database, the surface material recognition results of obstacles are obtained, and the weight of the LiDAR is adjusted according to the surface material recognition results of obstacles.

[0036] By adopting the above technical solution, the ambient light intensity is acquired through a visual camera, and the confidence weight of the visual sensor is dynamically adjusted according to the ambient light intensity. Specifically, when the ambient light intensity is greater than the light intensity threshold, the visual weight is adjusted to 0.3. By reducing the visual weight, the false detection of obstacles is reduced, enabling the ship to accurately identify obstacles and complete obstacle avoidance even in strong light scenarios. By using surface texture feature vectors to match the obstacle surface material recognition results in a preset material database, the confidence weight of the lidar is adjusted when the ship encounters transparent obstacles, which can prevent the ship from failing to identify transparent obstacles. This allows the unmanned system to dynamically allocate the confidence weight of each sensor according to the obstacle situation, further improving the ship's obstacle recognition accuracy and efficiency.

[0037] In a preferred embodiment, this application can be further configured as follows: the incremental obstacle avoidance strategy set is generated based on the preset reinforcement learning model according to the dynamic motion trajectory prediction parameters of the obstacle and the confidence weights of each sensor, specifically including:

[0038] Construct a state space for reinforcement learning, which includes relative distances to obstacles, collision times, and ship state data;

[0039] The action space is defined as a set of discrete obstacle avoidance strategies. Based on the near-end strategy optimization algorithm, the objective function is to minimize trajectory deviation and energy consumption, and a pre-set strategy network model is trained.

[0040] Based on trajectory prediction parameters and the confidence weights of each sensor, a set of obstacle avoidance strategies with priority ranking is generated in real time.

[0041] By employing the aforementioned technical solution, a state space for reinforcement learning is constructed. This state space includes key elements such as relative obstacle distances, collision times, and ship state data, ensuring that the reinforcement learning model can comprehensively perceive the current environmental state and providing a rich information foundation for subsequent decision-making. The introduction of relative obstacle distances and collision times enables the model to accurately assess potential collision risks, while the consideration of ship state data ensures that the obstacle avoidance strategy, while ensuring safety, also takes into account navigation efficiency and economy. The action space is defined as a discrete set of obstacle avoidance strategies, including strategies such as ship stopping and ship detouring, simplifying the decision-making process and enabling the model to quickly select the optimal solution from a limited set of strategies. Each strategy is optimized for a specific obstacle avoidance scenario, ensuring the effectiveness and flexibility of obstacle avoidance actions, based on proximal strategies. The optimization algorithm, with the objective function of minimizing trajectory deviation and energy consumption, trains a pre-defined policy network model. The application of a near-end policy optimization algorithm enables the model to continuously optimize policy performance while maintaining policy stability. By minimizing trajectory deviation and energy consumption, it ensures safe obstacle avoidance while minimizing deviations from the navigation path and energy waste, thus improving the efficiency and practicality of the obstacle avoidance strategy. Based on trajectory prediction parameters and the confidence weights of each sensor, a priority-ranked set of obstacle avoidance strategies is generated in real time. The fusion results of multi-source data are utilized to ensure the accuracy and reliability of the obstacle avoidance strategy. By comprehensively considering the confidence weights of each sensor, the risks and benefits of different strategies can be more accurately assessed, thereby generating a priority-ranked set of strategies, providing ships with more intelligent and flexible obstacle avoidance decision support.

[0042] In a preferred embodiment, this application can be further configured such that: based on the risk assessment level of the incremental obstacle avoidance strategy set, triggering the corresponding navigation control command to complete the ship's autonomous obstacle avoidance, specifically includes:

[0043] If the strategy priority is high risk level, trigger the ship stop command and activate the millimeter-wave radar omnidirectional scanning mode;

[0044] If the strategy priority is medium risk level, generate the ship's navigation detour trajectory and send it to the ship control system;

[0045] If the strategy priority is low risk, maintain the current route and periodically update obstacle status monitoring data.

[0046] By adopting the above technical solution, if the strategy priority is determined to be high-risk, a ship stop command is immediately triggered and the millimeter-wave radar omnidirectional scanning mode is activated, ensuring that the ship can stop quickly when facing extreme dangerous situations. At the same time, the high precision and all-weather operation capability of the millimeter-wave radar are used to conduct a comprehensive and blind-spot-free scan of the surrounding environment to obtain more detailed and accurate obstacle information. The ship stop not only provides valuable reaction time for the ship, but also provides more reliable data support for subsequent obstacle avoidance decisions.

[0047] If the strategy priority is medium risk level, a trajectory for the ship to detour is generated and sent to the ship control system for execution. This can cleverly avoid obstacles without changing the overall navigation direction of the ship. It not only avoids direct conflict between the ship and obstacles, but also ensures the continuity and smoothness of the navigation path, thereby improving the stability and safety of navigation.

[0048] If the strategy priority is low risk, the vessel will maintain its current course and periodically update obstacle status monitoring data. In a low-risk environment, the vessel does not need to perform large-scale obstacle avoidance maneuvers, but only needs to maintain continuous monitoring of the surrounding environment so that it can react quickly when necessary. Periodically updating obstacle status monitoring data ensures the vessel's real-time perception of the surrounding environment and provides strong data support for subsequent obstacle avoidance decisions.

[0049] Secondly, the above-mentioned inventive objective of this application is achieved through the following technical solutions:

[0050] A ship multi-source data interaction and processing system based on a Hub management platform, the system comprising:

[0051] The multi-source data acquisition module is used to acquire raw sensor data from the ship's lidar, visual camera, and millimeter-wave radar, and to preprocess the raw sensor data to generate an obstacle grid map with unified spatiotemporal reference.

[0052] The obstacle prediction module is used to perform multi-source data fusion on the obstacle grid map based on a dynamic Bayesian network to obtain dynamic motion trajectory prediction parameters of the obstacles.

[0053] The sensor weight adjustment module is used to acquire the ambient light intensity and obstacle surface material recognition results in real time, and dynamically allocate the confidence weight of each sensor according to the ambient light intensity and obstacle surface material recognition results.

[0054] The obstacle avoidance strategy generation module is used to generate an incremental obstacle avoidance strategy set based on a preset reinforcement learning model, according to the dynamic motion trajectory prediction parameters of the obstacle and the confidence weights of each sensor.

[0055] The control command generation module is used to trigger corresponding navigation control commands based on the risk assessment level of the incremental obstacle avoidance strategy set, so as to complete the ship's autonomous obstacle avoidance.

[0056] By adopting the above technical solution, during navigation, the ship simultaneously collects timestamps, spatial coordinates, and signal strength data through lidar, visual cameras, and millimeter-wave radar, ensuring the integrity of the original input of multi-source data. The solution performs lidar point cloud noise reduction, visual image semantic segmentation, and millimeter-wave radar signal analysis to extract dynamic obstacle features. A dynamic Bayesian network is used to fuse multi-source data and output obstacle trajectory prediction parameters, such as position, velocity, and acceleration. Based on ambient light intensity and material identification, the sensor confidence level is dynamically adjusted. Material identification, such as the reflectivity of glass surfaces, is introduced to correct the state transition matrix, improving the detection success rate of transparent obstacles. Then, a priority obstacle avoidance strategy set is generated by combining trajectory prediction and weight vectors through a reinforcement learning model. Based on the risk assessment level of the obstacle avoidance strategy, control commands are triggered to achieve obstacle avoidance control of the ship. The entire technical solution, through multi-sensor collaboration and dynamic weight adjustment, significantly improves the obstacle avoidance success rate in complex environments, such as strong light and transparent obstacles, constructing a complete closed loop for ship obstacle avoidance technology.

[0057] Thirdly, the above-mentioned objectives of this application are achieved through the following technical solutions:

[0058] An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described method for multi-source data interaction processing of ships based on a Hub management platform.

[0059] Fourthly, the above-mentioned objectives of this application are achieved through the following technical solutions:

[0060] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the above-described method for multi-source data interaction processing of ships based on a Hub management platform.

[0061] In summary, this application includes at least one of the following beneficial technical effects:

[0062] 1. During navigation, the ship simultaneously collects timestamps, spatial coordinates, and signal strength data via lidar, visual cameras, and millimeter-wave radar to ensure the integrity of the original input of multi-source data. This is achieved through lidar point cloud noise reduction, visual image semantic segmentation, and millimeter-wave radar signal analysis to extract dynamic obstacle features. A dynamic Bayesian network is then used to fuse the multi-source data, outputting obstacle trajectory prediction parameters such as position, velocity, and acceleration. Based on ambient light intensity and material identification, sensor confidence is dynamically adjusted. Material identification, such as the reflectivity of glass surfaces, is incorporated to correct the state transition matrix, improving the detection success rate of transparent obstacles. A reinforcement learning model, combined with trajectory prediction and weight vectors, generates a priority obstacle avoidance strategy set. Control commands are triggered based on the risk assessment level of the obstacle avoidance strategy, enabling obstacle avoidance control of the ship. This entire technical solution, through multi-sensor collaboration and dynamic weight adjustment, significantly improves the obstacle avoidance success rate in complex environments, such as strong light and transparent obstacles, constructing a complete closed loop for ship obstacle avoidance technology.

[0063] 2. By performing Gaussian filtering to reduce noise in the point cloud data of the lidar, the three-dimensional boundary feature point set of obstacles is extracted. By performing semantic segmentation on the images acquired by the visual camera, the category labels and surface texture features of obstacles are identified. By performing fast Fourier transform on the echo signal of the millimeter-wave radar, the radial velocity and distance information of obstacles are analyzed. Based on the real-time pose data of the ship, the data features output by various sensors are fused through spatiotemporal interpolation compensation, eliminating the spatiotemporal asynchrony and feature extraction incompatibility of multi-source sensor data, and generating a dynamic obstacle grid map in a unified coordinate system, so as to facilitate the subsequent generation of ship obstacle avoidance strategies and improve the accuracy of strategy generation.

[0064] 3. Based on the 10ms cycle of the lidar, the data from other sensors are aligned, and the visual camera data is interpolated to compensate for motion blur caused by insufficient visual frame rate. A 30fps continuous image stream is generated. The millimeter-wave radar data is subjected to sliding window mean filtering to eliminate millimeter-wave radar signal jitter and suppress speed change noise. The multi-source data is converted to the ship's body coordinate system to generate a dynamic obstacle grid map, thus solving the problem of asynchronous fusion of multi-source sensor data.

[0065] 4. By defining an obstacle state transition matrix, key parameters such as position, velocity, acceleration, and motion pattern are integrated, laying a solid foundation for subsequent trajectory prediction, improving prediction accuracy, and enhancing adaptability to complex obstacle motion patterns. The introduction of the state transition matrix allows for a more comprehensive understanding of the obstacle's dynamic characteristics, thereby reducing uncertainty in prediction. Using the LiDAR boundary feature point set as observation nodes, the initial state probability distribution of the obstacle is calculated. The high-precision ranging capability of the LiDAR ensures the reliability of the initial state estimation. The high-resolution characteristics of the LiDAR data make the extraction of boundary feature points particularly accurate, further improving the accuracy of the initial state probability distribution and providing strong data support for subsequent dynamic Bayesian network inference. The surface texture feature vector of the obstacle is fused to update the material. The relevant state transition probability weights enrich the descriptive dimensions of obstacle attributes, reflecting not only the physical characteristics of obstacles but also indirectly revealing their possible motion characteristics, such as sliding and rolling. By fusing this information, the dynamic Bayesian network can more intelligently adjust the state transition probabilities, making trajectory prediction more realistic. The trajectory prediction parameters of the dynamic Bayesian network are iteratively corrected based on radial velocity information, realizing real-time tracking and dynamic adjustment of the obstacle's motion state. Radial velocity, as a key indicator that directly reflects the relative motion state of the obstacle, enables the prediction model to quickly respond to changes in the obstacle's velocity and adjust the predicted trajectory in a timely manner, thereby effectively reducing prediction errors and improving the real-time performance and accuracy of trajectory prediction. By comprehensively considering various attribute information of obstacles, high-precision prediction of the dynamic motion trajectory of obstacles is achieved. Attached Figure Description

[0066] Figure 1 This is a flowchart of a method for multi-source data interaction processing of ships based on a Hub management platform in one embodiment of this application;

[0067] Figure 2 This is a flowchart illustrating the implementation of step S10 in a method for multi-source data interaction processing of ships based on a Hub management platform, according to an embodiment of this application.

[0068] Figure 3 This is a flowchart illustrating the implementation of step S14 in a method for multi-source data interaction processing of ships based on a Hub management platform, according to one embodiment of this application.

[0069] Figure 4 This is a flowchart illustrating the implementation of step S20 in a method for multi-source data interaction processing of ships based on a Hub management platform, as described in one embodiment of this application.

[0070] Figure 5 This is a flowchart illustrating the implementation of step S30 in a method for multi-source data interaction processing of ships based on a Hub management platform, as described in one embodiment of this application.

[0071] Figure 6 This is a flowchart illustrating the implementation of step S40 in a method for multi-source data interaction processing of ships based on a Hub management platform, as described in one embodiment of this application.

[0072] Figure 7 This is a flowchart illustrating the implementation of step S50 in a method for multi-source data interaction processing of ships based on a Hub management platform, as described in one embodiment of this application.

[0073] Figure 8 This is a principle block diagram of a ship multi-source data interaction processing system based on a Hub management platform according to one embodiment of this application;

[0074] Figure 9 This is a schematic diagram of an electronic device according to an embodiment of this application. Detailed Implementation

[0075] The present application will be further described in detail below with reference to the accompanying drawings.

[0076] In one embodiment, such as Figure 1 As shown, this application discloses a method for multi-source data interaction processing of ships based on a Hub management platform, which specifically includes the following steps:

[0077] S10: Acquire raw sensor data from the ship's lidar, vision camera, and millimeter-wave radar, preprocess the raw sensor data, and generate an obstacle grid map with unified spatiotemporal reference.

[0078] Specifically, by simultaneously acquiring timestamps, spatial coordinates, and signal strength data from LiDAR, visual cameras, and millimeter-wave radar, the integrity of the original input of multi-source data is ensured. Noise reduction of LiDAR point clouds, semantic segmentation of visual images, and analysis of millimeter-wave radar signals are performed to extract dynamic obstacle features. The advantages of different sensors are utilized: LiDAR provides high-precision distance information, visual cameras capture rich image details, and millimeter-wave radar maintains stable detection capabilities under adverse weather conditions. The preprocessing process ensures the spatiotemporal consistency of these data, laying a solid foundation for subsequent multi-source data fusion.

[0079] S20: Based on a dynamic Bayesian network, multi-source data fusion is performed on the obstacle grid map to obtain dynamic motion trajectory prediction parameters of the obstacles.

[0080] Specifically, by using dynamic Bayesian networks to fuse multi-source data, the system outputs obstacle trajectory prediction parameters such as position, velocity, and acceleration. Dynamic Bayesian networks can effectively handle uncertain information and improve the accuracy and robustness of obstacle trajectory prediction by fusing data from different sensors. This is crucial for ships to make timely and effective obstacle avoidance decisions in complex dynamic environments.

[0081] S30: Real-time acquisition of ambient light intensity and obstacle surface material recognition results, and dynamic allocation of confidence weights for each sensor based on the ambient light intensity and obstacle surface material recognition results.

[0082] Specifically, the impact of environmental factors on sensor performance was considered. For example, changes in lighting can affect the recognition performance of visual cameras, while the surface material of obstacles can affect the reflected signals of LiDAR. By dynamically adjusting the confidence weights, the advantages of each sensor can be fully utilized in different environments, thus improving overall reliability.

[0083] S40: Based on a preset reinforcement learning model, an incremental obstacle avoidance strategy set is generated according to the dynamic motion trajectory prediction parameters of the obstacle and the confidence weights of each sensor.

[0084] Specifically, reinforcement learning, through continuous learning and optimization, can generate optimal or suboptimal obstacle avoidance strategies based on the current environmental state. The generation of incremental strategy sets enables ships to quickly adjust when facing new obstacles or environmental changes, improving the flexibility and adaptability of obstacle avoidance decisions.

[0085] S50: Based on the risk assessment level of the incremental obstacle avoidance strategy set, trigger the corresponding navigation control command to complete the ship's autonomous obstacle avoidance.

[0086] Specifically, the introduction of risk assessment levels ensures that ships can take appropriate obstacle avoidance actions when facing different risk levels, which not only guarantees navigation safety but also improves navigation efficiency.

[0087] In this embodiment, during navigation, the ship simultaneously collects timestamps, spatial coordinates, and signal strength data via lidar, visual cameras, and millimeter-wave radar to ensure the integrity of the original input of multi-source data. Noise reduction of lidar point clouds, semantic segmentation of visual images, and analysis of millimeter-wave radar signals are performed to extract dynamic obstacle features. A dynamic Bayesian network is used to fuse multi-source data and output obstacle trajectory prediction parameters, such as position, velocity, and acceleration. Based on ambient light intensity and material identification, sensor confidence is dynamically adjusted. Material identification, such as the reflectivity of glass surfaces, is introduced to correct the state transition matrix, improving the detection success rate of transparent obstacles. A priority obstacle avoidance strategy set is generated using a reinforcement learning model combined with trajectory prediction and weight vectors. Control commands are triggered based on the risk assessment level of the obstacle avoidance strategy to achieve obstacle avoidance control of the ship. The entire technical solution, through multi-sensor collaboration and dynamic weight adjustment, significantly improves the obstacle avoidance success rate in complex environments, such as strong light and transparent obstacles, constructing a complete closed loop for ship obstacle avoidance technology.

[0088] In one embodiment, such as Figure 2As shown, in step S10, the raw sensor data is preprocessed to generate a spatiotemporally unified obstacle grid map, specifically including:

[0089] S11: Perform Gaussian filtering to reduce noise in the point cloud data of the lidar and extract the set of three-dimensional boundary feature points of the obstacle.

[0090] S12: Perform semantic segmentation on the images from the visual camera to identify obstacle category labels and surface texture feature vectors.

[0091] S13: Perform a fast Fourier transform on the echo signal from the millimeter-wave radar to analyze the radial velocity and distance information of the obstacle.

[0092] S14: Acquire the ship's real-time pose data, and based on the real-time pose data, the set of 3D boundary feature points of obstacles, obstacle category labels and surface texture feature vectors, and the radial velocity and distance information of obstacles, generate a dynamic obstacle grid map in a unified coordinate system through spatiotemporal interpolation compensation.

[0093] Specifically, by performing Gaussian filtering to reduce noise in the point cloud data of the lidar, the three-dimensional boundary feature point set of obstacles is extracted. By performing semantic segmentation on the images acquired by the visual camera, the category labels and surface texture features of obstacles are identified. By performing fast Fourier transform on the echo signal of the millimeter-wave radar, the radial velocity and distance information of obstacles are analyzed. Based on the real-time pose data of the ship, the data features output by various sensors are fused through spatiotemporal interpolation compensation, eliminating the spatiotemporal asynchrony and feature extraction incompatibility of multi-source sensor data, and generating a dynamic obstacle grid map in a unified coordinate system, which facilitates the subsequent generation of ship obstacle avoidance strategies and improves the accuracy of strategy generation.

[0094] In one embodiment, such as Figure 3 As shown, in step S14, the real-time pose data of the ship is acquired. Based on the real-time pose data, the three-dimensional boundary feature point set of obstacles, obstacle category labels and surface texture feature vectors, and the radial velocity and distance information of obstacles, a dynamic obstacle grid map in a unified coordinate system is generated through spatiotemporal interpolation compensation. Specifically, this includes:

[0095] S141: Establish a reference time axis based on the LiDAR data cycle, and perform interpolation compensation on the visual camera data to fill the time axis gaps.

[0096] S142: Perform sliding window mean filtering on millimeter-wave radar data to eliminate time asynchrony errors.

[0097] S143: Map the compensated data onto the 3D grid map corresponding to the ship's real-time pose data to generate a dynamic obstacle grid map.

[0098] Specifically, based on the 10ms cycle of the lidar, data from other sensors are aligned, and interpolation compensation is performed on the visual camera data to solve motion blur caused by insufficient visual frame rate, generating a 30fps continuous image stream. Sliding window mean filtering is applied to the millimeter-wave radar data to eliminate millimeter-wave radar signal jitter and suppress speed change noise. Multi-source data is converted to the ship's body coordinate system to generate a dynamic obstacle grid map, solving the problem of asynchronous fusion of multi-source sensor data.

[0099] In one embodiment, such as Figure 4 As shown, in step S20, multi-source data fusion is performed on the obstacle grid map based on a dynamic Bayesian network to obtain dynamic motion trajectory prediction parameters for the obstacles, specifically including:

[0100] S21: Define the obstacle state transition matrix, which includes position, velocity, acceleration and motion mode parameters.

[0101] Specifically, by defining an obstacle state transition matrix, key parameters such as position, velocity, acceleration, and motion pattern are integrated, laying a solid foundation for subsequent trajectory prediction, improving prediction accuracy, and enhancing adaptability to complex obstacle motion patterns. The introduction of the state transition matrix enables a more comprehensive understanding of the dynamic characteristics of obstacles, thereby reducing uncertainty in prediction.

[0102] S22: Using the set of boundary feature points of the lidar as observation nodes, calculate the probability distribution of the initial state of the obstacle.

[0103] Specifically, the boundary feature point set of the lidar is used as the observation node to calculate the probability distribution of the initial state of the obstacle. The high-precision ranging capability of lidar is utilized to ensure the reliability of the initial state estimation. The high-resolution characteristics of lidar data make the extraction of boundary feature points particularly accurate, thereby improving the accuracy of the initial state probability distribution and providing strong data support for subsequent dynamic Bayesian network inference.

[0104] S23: Fuse the surface texture feature vectors of obstacles and update the material-related state transition probability weights.

[0105] Specifically, by fusing the surface texture feature vectors of obstacles and updating the material-related state transition probability weights, the descriptive dimensions of obstacle attributes are enriched. This not only reflects the physical characteristics of obstacles but also indirectly reveals their possible motion characteristics, such as sliding and rolling. By fusing this information, the dynamic Bayesian network can more intelligently adjust the state transition probabilities, making trajectory prediction more realistic.

[0106] S24: Based on the radial velocity information, iteratively correct the trajectory prediction parameters of the dynamic Bayesian network.

[0107] Specifically, the trajectory prediction parameters of the dynamic Bayesian network are iteratively corrected based on radial velocity information, enabling real-time tracking and dynamic adjustment of the obstacle's motion state. Radial velocity, as a key indicator that directly reflects the relative motion state of the obstacle, allows the prediction model to respond quickly to changes in the obstacle's velocity and adjust the predicted trajectory in a timely manner, thereby effectively reducing prediction errors and improving the real-time performance and accuracy of trajectory prediction. By comprehensively considering various attribute information of the obstacle, high-precision prediction of the obstacle's dynamic motion trajectory is achieved.

[0108] In one embodiment, such as Figure 5 As shown, in step S30, the ambient light intensity and obstacle surface material recognition results are acquired in real time, and the confidence weights of each sensor are dynamically assigned based on the ambient light intensity and obstacle surface material recognition results. Specifically, this includes:

[0109] S31: Obtain ambient light intensity based on the visual camera, and dynamically adjust the weight of the visual sensor based on the ambient light intensity.

[0110] Specifically, the system acquires ambient light intensity through a visual camera and dynamically adjusts the confidence weight of the visual sensor based on the ambient light intensity. Specifically, when the ambient light intensity is greater than the light intensity threshold, the visual weight is adjusted to 0.3. By reducing the visual weight, the false detection of obstacles is reduced, enabling the ship to accurately identify obstacles and complete obstacle avoidance even in strong light scenarios.

[0111] S32: Based on the surface texture feature vector matching of the preset material database, obtain the obstacle surface material recognition result, and adjust the weight of the LiDAR according to the obstacle surface material recognition result.

[0112] Specifically, by using surface texture feature vectors in a pre-set material database, the surface material recognition results of obstacles are matched. When a ship encounters a transparent obstacle, the confidence weight of the LiDAR is adjusted, which can prevent the ship from failing to recognize transparent obstacles. This allows the unmanned system to dynamically allocate the confidence weight of each sensor according to the obstacle situation, further improving the ship's accuracy and efficiency in identifying obstacles.

[0113] In one embodiment, such as Figure 6 As shown, in step S40, based on a preset reinforcement learning model, an incremental obstacle avoidance strategy set is generated according to the dynamic trajectory prediction parameters of the obstacle and the confidence weights of each sensor, specifically including:

[0114] S41: Construct a state space for reinforcement learning, which includes relative distances to obstacles, collision times, and ship state data.

[0115] Specifically, a state space for reinforcement learning is constructed, which includes key elements such as relative distances to obstacles, collision times, and ship state data. This ensures that the reinforcement learning model can fully perceive the current environmental state, providing a rich information foundation for subsequent decision-making. The introduction of relative distances to obstacles and collision times enables the model to accurately assess potential collision risks, while the consideration of ship state data ensures that the obstacle avoidance strategy takes into account both safety and navigation efficiency and economy.

[0116] S42: Define the action space as a set of discrete obstacle avoidance strategies. Based on the near-end strategy optimization algorithm, with the objective function of minimizing trajectory deviation and energy consumption, train the preset strategy network model.

[0117] Specifically, the action space is defined as a set of discrete obstacle avoidance strategies, including strategies such as stopping the ship and navigating around obstacles. This simplifies the decision-making process, enabling the model to quickly select the optimal solution from a finite set of strategies. Each strategy is optimized for a specific obstacle avoidance scenario, ensuring the effectiveness and flexibility of obstacle avoidance actions.

[0118] Furthermore, based on the near-end policy optimization algorithm, a pre-defined policy network model is trained with the objective function of minimizing trajectory deviation and energy consumption. The application of the near-end policy optimization algorithm enables the model to continuously optimize policy performance while maintaining policy stability. By minimizing trajectory deviation and energy consumption, the model can minimize deviations from the navigation path and energy waste while ensuring safe obstacle avoidance, thereby improving the efficiency and practicality of the obstacle avoidance strategy.

[0119] S43: Generate a set of obstacle avoidance strategies in real time, prioritizing them according to the trajectory prediction parameters and the confidence weights of each sensor.

[0120] Specifically, based on trajectory prediction parameters and the confidence weights of each sensor, a priority-ranked set of obstacle avoidance strategies is generated in real time. By utilizing the fusion results of multi-source data, the accuracy and reliability of the obstacle avoidance strategies are ensured. By comprehensively considering the confidence weights of each sensor, the risks and benefits of different strategies can be assessed more accurately, thereby generating a priority-ranked set of strategies and providing ships with more intelligent and flexible obstacle avoidance decision support.

[0121] In one embodiment, such as Figure 7 As shown, in step S50, based on the risk assessment level of the incremental obstacle avoidance strategy set, the corresponding navigation control command is triggered to complete the ship's autonomous obstacle avoidance, specifically including:

[0122] S51: If the strategy priority is high risk level, trigger the ship stop command and activate the millimeter wave radar omnidirectional scanning mode.

[0123] Specifically, if the strategy priority is determined to be high-risk (TTC < 2 seconds), a ship stop command is immediately triggered, and the millimeter-wave radar omnidirectional scanning mode is activated. This ensures that the ship can quickly stop when facing extremely dangerous situations. At the same time, the high precision and all-weather operation capability of the millimeter-wave radar are used to conduct a comprehensive and blind-spot-free scan of the surrounding environment to obtain more detailed and accurate obstacle information. The ship stop not only provides valuable reaction time for the ship, but also provides more reliable data support for subsequent obstacle avoidance decisions.

[0124] S52: If the strategy priority is medium risk level, generate the ship's detour trajectory and send it to the ship control system.

[0125] Specifically, if the strategy priority is medium risk level (2 seconds ≤ TTC ≤ 5 seconds), a trajectory for the ship to detour is generated and sent to the ship control system for execution. This can cleverly avoid obstacles without changing the overall navigation direction of the ship. It not only avoids direct conflict between the ship and the obstacle, but also ensures the continuity and smoothness of the navigation path, thereby improving the stability and safety of navigation.

[0126] S53: If the strategy priority is low risk level, maintain the current route and periodically update obstacle status monitoring data.

[0127] Specifically, if the strategy priority is low risk (TTC > 5 seconds), the vessel will maintain its current course and periodically update obstacle status monitoring data. In a low-risk environment, the vessel does not need to perform large-scale obstacle avoidance maneuvers, but only needs to maintain continuous monitoring of the surrounding environment so that it can react quickly when necessary. Periodically updating obstacle status monitoring data ensures the vessel's real-time perception of the surrounding environment and provides strong data support for subsequent obstacle avoidance decisions.

[0128] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0129] In one embodiment, a ship multi-source data interaction processing system based on a Hub management platform is provided. This system corresponds one-to-one with the ship multi-source data interaction processing method based on a Hub management platform described in the above embodiments. Figure 8 As shown, this ship multi-source data interaction and processing system based on the Hub management platform includes a multi-source data acquisition module, an obstacle prediction module, a sensor weight adjustment module, an obstacle avoidance strategy generation module, and a control command generation module. Detailed descriptions of each functional module are as follows:

[0130] The multi-source data acquisition module is used to acquire raw sensor data from the ship's lidar, visual camera, and millimeter-wave radar, and to preprocess the raw sensor data to generate an obstacle grid map with unified spatiotemporal reference.

[0131] The obstacle prediction module is used to perform multi-source data fusion on the obstacle grid map based on a dynamic Bayesian network to obtain dynamic motion trajectory prediction parameters of the obstacles.

[0132] The sensor weight adjustment module is used to acquire the ambient light intensity and obstacle surface material recognition results in real time, and dynamically allocate the confidence weight of each sensor according to the ambient light intensity and obstacle surface material recognition results.

[0133] The obstacle avoidance strategy generation module is used to generate an incremental obstacle avoidance strategy set based on a preset reinforcement learning model, according to the dynamic motion trajectory prediction parameters of the obstacle and the confidence weights of each sensor.

[0134] The control command generation module is used to trigger corresponding navigation control commands based on the risk assessment level of the incremental obstacle avoidance strategy set, so as to complete the ship's autonomous obstacle avoidance.

[0135] Specific limitations regarding the multi-source data interaction processing system for ships based on the Hub management platform can be found in the limitations of the multi-source data interaction processing method for ships based on the Hub management platform mentioned above, and will not be repeated here. Each module in the aforementioned multi-source data interaction processing system for ships based on the Hub management platform can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in the electronic device, or stored in the memory of the electronic device as software, so that the processor can call and execute the corresponding operations of each module.

[0136] In one embodiment, an electronic device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 9 As shown, the electronic device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database stores monitoring data from multiple patients and diabetes analysis models. The network interface communicates with external terminals via a network connection. When executed by the processor, the computer program implements a method for multi-source data interaction processing on a ship based on a Hub management platform.

[0137] In one embodiment, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to perform the following steps:

[0138] The raw sensor data from the lidar, vision camera and millimeter-wave radar on board the ship are acquired, and the raw sensor data is preprocessed to generate an obstacle grid map with unified spatiotemporal reference.

[0139] Based on a dynamic Bayesian network, multi-source data fusion is performed on the obstacle grid map to obtain dynamic motion trajectory prediction parameters for the obstacles.

[0140] The system acquires real-time ambient light intensity and obstacle surface material recognition results, and dynamically assigns confidence weights to each sensor based on these results.

[0141] Based on a preset reinforcement learning model, an incremental obstacle avoidance strategy set is generated according to the dynamic motion trajectory prediction parameters of the obstacle and the confidence weight of each sensor.

[0142] Based on the risk assessment level of the incremental obstacle avoidance strategy set, the corresponding navigation control command is triggered to complete the ship's autonomous obstacle avoidance.

[0143] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor:

[0144] The raw sensor data from the lidar, vision camera and millimeter-wave radar on board the ship are acquired, and the raw sensor data is preprocessed to generate an obstacle grid map with unified spatiotemporal reference.

[0145] Based on a dynamic Bayesian network, multi-source data fusion is performed on the obstacle grid map to obtain dynamic motion trajectory prediction parameters for the obstacles.

[0146] The system acquires real-time ambient light intensity and obstacle surface material recognition results, and dynamically assigns confidence weights to each sensor based on these results.

[0147] Based on a preset reinforcement learning model, an incremental obstacle avoidance strategy set is generated according to the dynamic motion trajectory prediction parameters of the obstacle and the confidence weight of each sensor.

[0148] Based on the risk assessment level of the incremental obstacle avoidance strategy set, the corresponding navigation control command is triggered to complete the ship's autonomous obstacle avoidance.

[0149] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0150] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0151] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A method for multi-source data interaction processing of ships based on a Hub management platform, characterized in that, The method for multi-source data interaction processing of ships based on the Hub management platform includes the following steps: The raw sensor data from the lidar, vision camera and millimeter-wave radar on board the ship are acquired, and the raw sensor data is preprocessed to generate an obstacle grid map with unified spatiotemporal reference. Based on a dynamic Bayesian network, multi-source data fusion is performed on the obstacle grid map to obtain dynamic motion trajectory prediction parameters for the obstacles. The system acquires real-time ambient light intensity and obstacle surface material recognition results, and dynamically assigns confidence weights to each sensor based on these results. Based on a preset reinforcement learning model, an incremental obstacle avoidance strategy set is generated according to the dynamic motion trajectory prediction parameters of the obstacle and the confidence weight of each sensor. Based on the risk assessment level of the incremental obstacle avoidance strategy set, the corresponding navigation control command is triggered to complete the ship's autonomous obstacle avoidance.

2. The method for multi-source data interaction processing of ships based on a Hub management platform according to claim 1, characterized in that, The preprocessing of the raw sensor data to generate a spatiotemporally unified obstacle grid map specifically includes: Gaussian filtering is applied to the point cloud data of the lidar to reduce noise and extract the three-dimensional boundary feature point set of the obstacle; Perform semantic segmentation on images from a visual camera to identify obstacle category labels and surface texture feature vectors; Fast Fourier transform is performed on the echo signal of millimeter-wave radar to analyze the radial velocity and distance information of obstacles; The system acquires real-time pose data of the ship and generates a dynamic obstacle grid map in a unified coordinate system based on the real-time pose data, the three-dimensional boundary feature point set of obstacles, the obstacle category labels and surface texture feature vectors, the radial velocity and distance information of obstacles, and spatiotemporal interpolation compensation.

3. The method for multi-source data interaction processing of ships based on a Hub management platform according to claim 2, characterized in that, The acquisition of the ship's real-time pose data, based on the real-time pose data, the set of three-dimensional boundary feature points of obstacles, obstacle category labels and surface texture feature vectors, and the radial velocity and distance information of obstacles, generates a dynamic obstacle grid map in a unified coordinate system through spatiotemporal interpolation compensation, specifically including: A baseline time axis is established based on the LiDAR data cycle, and interpolation compensation is performed on the visual camera data to fill the time axis gaps. Sliding window mean filtering is applied to millimeter-wave radar data to eliminate time asynchrony errors; The compensated data is mapped onto a 3D grid map corresponding to the ship's real-time pose data to generate a dynamic obstacle grid map.

4. The method for multi-source data interaction processing of ships based on a Hub management platform according to claim 1, characterized in that, The process of fusing multi-source data from the obstacle grid map using a dynamic Bayesian network to obtain dynamic motion trajectory prediction parameters for the obstacles specifically includes: Define an obstacle state transition matrix, which includes position, velocity, acceleration, and motion mode parameters; Using the set of boundary feature points of the lidar as observation nodes, the probability distribution of the initial state of the obstacle is calculated. Fuse the surface texture feature vectors of obstacles and update the material-related state transition probability weights; Based on the radial velocity information, the trajectory prediction parameters of the dynamic Bayesian network are iteratively corrected.

5. The method for multi-source data interaction processing of ships based on a Hub management platform according to claim 1, characterized in that, The process of acquiring real-time ambient light intensity and obstacle surface material recognition results, and dynamically allocating confidence weights for each sensor based on these results, specifically includes: The ambient light intensity is obtained based on the visual camera, and the weights of the visual sensor are dynamically adjusted based on the ambient light intensity. Based on surface texture feature vector matching with a pre-set material database, the surface material recognition results of obstacles are obtained, and the weight of the LiDAR is adjusted according to the surface material recognition results of obstacles.

6. The method for multi-source data interaction processing of ships based on a Hub management platform according to claim 1, characterized in that, The preset reinforcement learning model generates an incremental obstacle avoidance strategy set based on the dynamic trajectory prediction parameters of the obstacle and the confidence weights of each sensor, specifically including: Construct a state space for reinforcement learning, which includes relative distances to obstacles, collision times, and ship state data; The action space is defined as a set of discrete obstacle avoidance strategies. Based on the near-end strategy optimization algorithm, the objective function is to minimize trajectory deviation and energy consumption, and a pre-set strategy network model is trained. Based on trajectory prediction parameters and the confidence weights of each sensor, a set of obstacle avoidance strategies with priority ranking is generated in real time.

7. The method for multi-source data interaction processing of ships based on a Hub management platform according to claim 1, characterized in that, The step of triggering corresponding navigation control commands based on the risk assessment level of the incremental obstacle avoidance strategy set to complete autonomous obstacle avoidance of the vessel specifically includes: If the strategy priority is high risk level, trigger the ship stop command and activate the millimeter-wave radar omnidirectional scanning mode; If the strategy priority is medium risk level, generate the ship's navigation detour trajectory and send it to the ship control system; If the strategy priority is low risk, maintain the current route and periodically update obstacle status monitoring data.

8. A ship multi-source data interaction and processing system based on a Hub management platform, characterized in that, The ship multi-source data interaction and processing system based on the Hub management platform includes: The multi-source data acquisition module is used to acquire raw sensor data from the ship's lidar, visual camera, and millimeter-wave radar, and to preprocess the raw sensor data to generate an obstacle grid map with unified spatiotemporal reference. The obstacle prediction module is used to perform multi-source data fusion on the obstacle grid map based on a dynamic Bayesian network to obtain dynamic motion trajectory prediction parameters of the obstacles. The sensor weight adjustment module is used to acquire the ambient light intensity and obstacle surface material recognition results in real time, and dynamically allocate the confidence weight of each sensor according to the ambient light intensity and obstacle surface material recognition results. The obstacle avoidance strategy generation module is used to generate an incremental obstacle avoidance strategy set based on a preset reinforcement learning model, according to the dynamic motion trajectory prediction parameters of the obstacle and the confidence weights of each sensor. The control command generation module is used to trigger corresponding navigation control commands based on the risk assessment level of the incremental obstacle avoidance strategy set, so as to complete the ship's autonomous obstacle avoidance.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the ship multi-source data interaction processing method based on the Hub management platform as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the ship multi-source data interaction processing method based on the Hub management platform as described in any one of claims 1 to 7.

Images