A marine ship navigation risk grading intelligent early warning method based on multi-source spatio-temporal data fusion

Abstract

The application discloses a kind of based on multi-source spatio-temporal data fusion's sea ship navigation risk grading intelligent early warning method, it is related to sea navigation safety technical field.This method first constructs the unified space-time index of 1km×1km space grid and 15 minutes time window, six kinds of heterogeneous navigation data are converted into standardized risk feature vector in combination with three-level data credibility weighted fusion mechanism;Further establish four-dimensional dynamic weight evaluation model, introduce ship anti-risk ability correction coefficient calculation single ship exclusive actual navigation risk value, based on inertial trajectory predicts future multi-period peak risk;Through five kinds of core trigger condition realizes grading early warning, finally according to risk grade and ship attribute generates differentiated disposal suggestion, synchronously to multiple terminal realizes ship-shore cooperation.The risk assessment accuracy of the application reaches 91.4%, can perceive risk in advance 1-6 hours, effectively improves early warning pertinence and disposal efficiency, and comprehensively guarantees sea navigation safety.

Description

Technical Field

[0001] This invention relates to the field of maritime navigation safety and intelligent early warning technology, and in particular to an intelligent early warning method for maritime vessel navigation risk classification based on multi-source spatiotemporal data fusion. Background Technology

[0002] With the rapid development of the global shipping industry, the number of ships at sea continues to increase, and the navigation environment is becoming increasingly complex. Traditional ship route planning and risk warning technologies are no longer able to meet the needs of modern shipping safety and efficiency.

[0003] Currently, the mainstream methods for early warning of ship navigation risks have three main shortcomings: One issue is the lack of a single risk warning dimension. For example, the ship navigation warning method disclosed in patent CN116312062A only integrates radar and navigation data to warn of the risk of collision between routes, without covering other core risk dimensions such as meteorology, geography, and the ship's own condition. Another example is patent CN108313236A, which only makes simple threshold judgments on ship operating parameters and basic environmental data, and cannot comprehensively consider the combined effects of multiple factors.

[0004] Secondly, the depth of data fusion is insufficient. For example, the ship navigation early warning method based on geofencing disclosed in patent CN120932501A collects navigation parameters from multiple sources and sets basic weights, but only performs direct weighted splicing at the data level. It does not establish a unified spatiotemporal index to achieve deep correlation, nor does it consider the reliability differences of different data sources.

[0005] Thirdly, the early warning is highly homogenized. For example, the ship navigation early warning method based on potential field disclosed in patent CN121747368A constructs a multi-dimensional potential field to calculate the risk index, but it adopts a uniform risk threshold and assessment standard for all ships without modifying it based on the ship's own attributes such as tonnage, wind resistance, and cargo hazard level, and it cannot generate differentiated handling suggestions.

[0006] Therefore, there is an urgent need to develop a marine intelligent early warning technology that can achieve deep spatiotemporal fusion of multi-source data, multi-dimensional personalized risk comprehensive assessment, early prediction of future trajectories, and graded differentiated handling guidance to address the shortcomings of the existing technologies. Summary of the Invention

[0007] The purpose of this invention is to address the aforementioned shortcomings of existing technologies by providing a method for intelligent early warning of maritime navigation risks based on multi-source spatiotemporal data fusion. By constructing a spatiotemporally correlated multi-source data fusion architecture, a multi-dimensional comprehensive risk assessment model, and a tiered personalized early warning system, this invention systematically solves problems in existing technologies such as insufficient data fusion depth, single risk assessment dimensions, homogeneous and inefficient early warnings, and poor ship-shore coordination. This enables early perception, accurate identification, tiered early warning, and collaborative handling of maritime navigation risks, comprehensively improving the level of maritime navigation safety.

[0008] To achieve the above objectives, the present invention adopts the following technical solution: A method for intelligent early warning of maritime vessel navigation risk classification based on multi-source spatiotemporal data fusion includes the following steps: S1 synchronously collects six types of heterogeneous navigation raw data: satellite navigation, AIS ship dynamics, marine meteorology, tides and currents, electronic nautical charts, and port traffic density. All raw data are preprocessed and standardized to obtain standardized basic data. S2 constructs a unified spatiotemporal index, and based on a three-level data credibility weighted fusion mechanism, performs spatiotemporal correlation fusion on standardized basic data to extract four types of standardized risk feature vectors: environmental risk features, traffic risk features, ship risk features, and geographical risk features. S3 configures the basic weights of multi-dimensional risk factors, dynamically adjusts the basic weights in combination with navigation scenarios and ship attributes, and calculates the basic risk values ​​of each spatiotemporal grid within the ship monitoring range based on the adjusted weights. S4 introduces a ship risk resistance capability correction coefficient to correct the basic risk value, obtains the actual navigation risk value specific to a single ship, and predicts the peak risk for multiple future periods based on the inertial trajectory corrected by the channel boundary. S5 sets up multiple core risk trigger conditions, combines actual navigation risk values ​​with predicted peak risks to conduct graded early warning judgments, generates differentiated handling suggestions based on risk level, risk type and vessel attributes, and pushes them to multiple terminals simultaneously to achieve ship-shore collaborative handling.

[0009] Furthermore, the preprocessing and standardization processes in step S1 include timestamp synchronization, coordinate system unification, missing value handling, and outlier removal; among which, missing value handling employs inverse distance weighted interpolation based on spatiotemporal proximity for completion, and the interpolation calculation formula is: ; in, For interpolation point values, Given the point value, The spatiotemporal distance between the known point and the interpolation point is given by k=2, where k=2 is the power exponent. Outlier removal uses the 3σ criterion: for continuous data of the same type within the same spatiotemporal grid, if data point x satisfies |x If μ|>3σ, it is considered an outlier and removed, where μ is the data mean and σ is the data standard deviation.

[0010] Furthermore, the three-level data credibility weighted fusion mechanism divides data into three levels, A, B, and C, based on source reliability, update frequency, and completeness, with corresponding weights of 1.0, 0.85, and 0.70, respectively. Level A corresponds to official authoritative sources with an update frequency of no more than 1 hour and a data integrity rate of no less than 95%; Level B corresponds to commercial data sources with an update frequency of no more than 6 hours and a data integrity rate of no less than 80%; and Level C corresponds to third-party shared data with an update frequency of no more than 24 hours and a data integrity rate of no less than 60%.

[0011] Furthermore, the unified spatiotemporal index adopts a 1km×1km spatial grid and a 15-minute time window; during spatiotemporal correlation weighted fusion, the multi-source features within the same spatiotemporal grid are weighted and averaged according to the data credibility weights to output the unified grid risk basic features. The fusion calculation formula is the sum of the products of each dimension feature value and the corresponding credibility weight divided by the sum of the total weights.

[0012] Furthermore, the four types of standardized risk feature vectors are calculated as follows: Environmental risk characteristics: =0.4× +0.3× +0.2× +0.1× ; Among them, wind speed characteristics =min(V / 30,1), where V is the wind speed in m / s; significant wave height characteristic. =min(H / 10,1), where H is the wave height in meters; visibility characteristic. =1-min(S / 10,1), where S is visibility in nautical miles; rainfall intensity characteristics =min(R / 50,1), where R is the hourly rainfall in mm; Traffic risk characteristics: =min(N / 10,1), where N is the number of ships within a 3-nautical-mile radius; The inherent risk characteristics of the vessel: =0.4×tonnage+0.3×draft+0.3×cargo; Among them, tonnage characteristics =min(T / 100000,1), where T is the ship's deadweight tons; draft characteristics =min(D / 15,1), where D is the ship's draft in meters; cargo hazard rating characteristics. =C / 5, where C is the IMDG hazard class of the cargo and its value ranges from 1 to 5; Geographic risk characteristics: =0.6× +0.4× ; Among them, water depth characteristics =1-min(WD / 20,1), where WD is the current water depth in meters; characteristics of historical accident high-incidence areas. =1, the value is 1 when it is located in a high-incidence area, and 0 otherwise.

[0013] Furthermore, the basic weights of the risk factors are: environmental risk 0.4, traffic risk 0.3, ship's own risk 0.2, and geographical risk 0.1; the dynamic weight adjustment formula is as follows: = ×(1+ ×α), after adjustment, all factor weights are normalized to ensure the sum of weights is 1; where The basic weights of the i-th type of risk factor, The corresponding adjustment coefficient is α, which is the scene association parameter.

[0014] Furthermore, the formula for calculating the basic risk value is as follows: =( × + × + × + × )×100; The formula for calculating the actual navigation risk value is as follows: = Where T is the data reliability and its value ranges from 0 to 1, and C is the ship's risk resistance coefficient and its value ranges from 0.35 to 1.0; The formula for calculating the ship's risk resistance coefficient is C = 0.35 + 0.65 × (0.4 × ... +0.3× +0.3× ),in This is the wind resistance rating coefficient. For the ship's stability coefficient, For ship handling performance coefficient; The formula for calculating the risk value of future trajectory prediction is as follows: =0.5× +0.3× +0.2× ,in , , These are the peak risk values ​​for the next 1 hour, 3 hours, and 6 hours, respectively.

[0015] Furthermore, the multiple core risk triggering conditions include five categories, and the early warning process will be initiated if any one of them is met: Category 1: The comprehensive risk value of the vessel in the navigation area within the next hour exceeds the medium risk threshold by 30 points. The second category is when the typhoon's path deviates beyond the preset threshold, and the typhoon's 7-level wind circle will cover the shipping area within 6 hours. Category 3: There are no fewer than 5 vessels within 3 nautical miles of the vessel, and any two vessels meet the condition that the closest encounter distance is no more than 0.5 nautical miles and the closest encounter time is no more than 10 minutes. The fourth category involves malfunctions in the ship's main engine, steering gear, or navigation equipment, which, when combined with the ship's own risk value by 20 points, exceeds the medium risk threshold. The fifth category is vessels that have entered areas with a high incidence of historical accidents, and the cosine similarity between the current sea state and the sea state at the time of the accident is not less than 0.7.

[0016] Furthermore, the tiered early warning system classifies navigation risks into four levels: Low risk corresponds to a comprehensive risk score of 0 to 30, with no early warning and the risk situation updated every 30 minutes; Medium risk corresponds to a comprehensive risk score of 31 to 60, and a yellow alert is issued. Text prompts are given through the shipboard terminal, and risk details are pushed out every 10 minutes. High risk corresponds to a comprehensive risk value of 61 to 80 points, an orange alert is issued, the shipborne terminal activates an audible and visual alarm, pushes risk details and preliminary handling suggestions every 5 minutes, and automatically notifies the shore-based dispatch center; Extremely high risk corresponds to a comprehensive risk value of 81 to 100 points, triggering a red alert. The shipborne terminal will activate an emergency audible and visual alarm, continuously push risk details and emergency response suggestions, and automatically notify the maritime regulatory authorities and the maritime search and rescue center.

[0017] Furthermore, the differentiated handling recommendations are generated based on the navigation safety rule base and emergency response plan, combined with the ship type, tonnage, wind resistance level, cargo hazard level and crew attributes; multi-terminal synchronous early warning is generated through 5G and satellite communication links, and the early warning information is pushed to the shipborne navigation terminal, crew smart wearable devices and shore-based dispatch platform to achieve ship-shore emergency communication and data synchronous retention.

[0018] Compared with the prior art, the present invention has the following beneficial effects: 1. Improved the comprehensiveness and accuracy of risk assessment. By constructing a spatiotemporal grid index and a multi-dimensional dynamic weight fusion mechanism, deep fusion of six types of heterogeneous navigation data was achieved. This enables comprehensive consideration of multi-dimensional risk factors such as environment, traffic, and the ship itself, resulting in a risk assessment accuracy rate of 91.4%, which is more than 35% higher than traditional methods.

[0019] 2. It enables early risk perception and tiered warning. It can detect navigation risks 1-6 hours in advance and issue tiered warnings based on the severity of the risks, allowing sufficient time for crew and shore-based personnel to respond and effectively reducing the accident rate.

[0020] 3. Enhanced the targeting and effectiveness of early warnings. A personalized response suggestion generation system was established, capable of generating differentiated response suggestions based on the actual conditions of different vessels and crew members. This effectively avoids crew fatigue caused by homogeneous alarms, and improves early warning response speed by 50%.

[0021] 4. Achieved efficient ship-shore collaboration. Through multi-terminal information synchronization links and emergency communication channels, real-time information exchange between ship and shore was ensured, emergency response time was shortened by 30%, and the overall safety level of maritime navigation was significantly improved. Attached Figure Description

[0022] The present invention will be further described below with reference to the accompanying drawings and embodiments: Figure 1 This is a schematic diagram of the overall architecture of the method of the present invention; Figure 2 This is a schematic diagram of the data preprocessing and credibility weighting logic of the present invention; Figure 3 This is a schematic diagram illustrating the establishment of the spatiotemporal grid index of the present invention; Figure 4 This is a schematic diagram of the same-grid multi-source credibility weighted fusion logic of the present invention; Figure 5 This is a schematic diagram illustrating the weighted and normalized calculation of the four types of risk factors in this invention; Figure 6 This is a schematic diagram illustrating the logic of the comprehensive risk score and ship capability correction in this invention; Figure 7 This is a schematic diagram of the early warning triggering logic of the present invention. Detailed Implementation

[0023] The specific embodiments of the present invention will be further described in detail with reference to the accompanying drawings.

[0024] See Figure 1The intelligent early warning method for maritime vessel navigation risk classification based on multi-source spatiotemporal data fusion described in this invention has an overall architecture consisting of three core parts: multi-source spatiotemporal data fusion and risk feature extraction, multi-dimensional comprehensive risk assessment, and graded personalized early warning and ship-shore collaborative handling. In this embodiment, the shipborne terminal adopts an industrial-grade embedded navigation terminal, supporting dual-mode positioning of Beidou-3 and GPS, with a positioning sampling frequency of 1Hz; the AIS receiving module adopts Class A shipborne equipment compliant with IMO standards, with a dynamic data update cycle of no more than 5 minutes; marine meteorological and tidal current data are obtained from the official data center of the State Oceanic Administration through a maritime satellite communication link, with a spatial resolution of 1km×1km and a temporal resolution of 1 hour for meteorological data, and a temporal resolution of 15 minutes for tidal current data; the electronic nautical chart adopts an official electronic nautical chart compliant with IHOS-57 standards; port traffic density data is obtained through the interface of the Maritime Safety Administration's VTS system, with an update frequency of 5 minutes; the shore-based dispatch platform is deployed at the maritime supervision center, using dual-channel communication of 5G public network and maritime satellite to ensure data transmission continuity. When the system is running, it first collects six types of core navigation risk-related data synchronously through standardized API interfaces, satellite communication links and shipborne sensors. For specific data types and sampling parameters, please refer to Table 1.

[0025] Table 1. Explanation of Core Risk Data Collection

[0026] After data collection, the raw data undergoes preprocessing and standardization. The processing logic is detailed below. Figure 2 Specific preprocessing rules are shown in Table 2. Preprocessing first converts all timestamps to UTC standard time and all location data to the WGS-84 coordinate system to unify time and spatial references. For missing data, inverse distance weighted interpolation based on spatiotemporal proximity is used for completion. The missing data determination rule is that if no valid data of the same spatiotemporal grid and data type is collected within the current time window, it is considered a missing value. The interpolation calculation is centered on the missing point, selecting spatiotemporally adjacent valid data points, and calculating the interpolation result using the IDW formula, with the exponent k set to 2. If the same data type is continuously missing for more than 30 minutes, or the missing rate within a single time window is greater than 50%, the data segment is deemed unusable, no interpolation is performed, and the data reliability level of the corresponding area is downgraded by one level: A to B, B to C, and C remains unchanged. For outlier data, the 3σ criterion is used for outlier removal. For continuous data of the same type within the same spatiotemporal grid, the mean μ and standard deviation σ are calculated. If data point x satisfies |x If μ| > 3σ, it is considered an outlier and removed. For extreme weather and sea state data that conform to physical laws, such as typhoons and giant waves, even if they exceed the 3σ range, they are still retained and marked as high-risk data. After outlier removal, the spatiotemporal proximity interpolation method is used to complete the data. If there are consecutive outliers or the outlier rate is too high within the same grid, the data in that area is marked as unreliable and the reliability level is downgraded.

[0027] Table 2 Preprocessing Logic Description

[0028] Data credibility is graded into three levels based on the reliability of the data source, update frequency, and completeness. Level A corresponds to official and authoritative sources with an update frequency of no more than 1 hour and a data completeness rate of no less than 95%, with a weight of 1.0; Level B corresponds to commercial data sources with an update frequency of no more than 6 hours and a data completeness rate of no less than 80%, with a weight of 0.85; and Level C corresponds to third-party shared data with an update frequency of no more than 24 hours and a data completeness rate of no less than 60%, with a weight of 0.70.

[0029] After preprocessing, a unified spatiotemporal index is constructed to achieve spatiotemporal correlation and fusion of data. For details on how to construct the spatiotemporal index, please refer to [link / reference needed]. Figure 3 A 1km×1km spatial grid and a 15-minute time window were used to divide the target sea area into continuous spatiotemporal units, and all standardized data were mapped to the corresponding spatiotemporal grid units. Subsequently, multi-source feature data within the same spatiotemporal grid were fused using a reliability-weighted method; the fusion logic is described in [link to fusion logic]. Figure 4 During the fusion calculation, the feature vectors of each data source are multiplied by their corresponding confidence weights, summed, and then divided by the sum of the total weights to obtain the unified risk basic features of the grid.

[0030] Based on the fused grid-based data, four types of standardized risk feature vectors were extracted, with specific calculation rules shown in Table 3. Environmental risk features were obtained by weighting four indicators: wind speed, significant wave height, visibility, and rainfall intensity, with weights of 0.4, 0.3, 0.2, and 0.1, respectively. Wind speed was normalized to a maximum value of 30 m / s, significant wave height to a maximum value of 10 m, visibility to a maximum value of 10 nautical miles (inverse normalization), and rainfall intensity to a maximum value of 50 mm / hour. Traffic risk features were based on the number of vessels within a 3-nautical-mile radius, normalized to a maximum value of 10 vessels. Vessel risk features were obtained by weighting three indicators: tonnage, draft, and cargo hazard class, with weights of 0.4, 0.3, and 0.3, respectively. Tonnage was normalized to a maximum value of 100,000 deadweight tons, draft to a maximum value of 15 m, and cargo hazard class to IMDG standards 1 to 5. The geographical risk characteristics are obtained by weighting two indicators: water depth and historical high-incidence areas, with weights of 0.6 and 0.4, respectively. The water depth is back-normalized with 20m as the full value. The accident characteristic value is 1 when the location is in a historical high-incidence area, and 0 otherwise.

[0031] Table 3. Rules for Calculating Risk Feature Vectors

[0032] After feature extraction is completed, the multi-dimensional risk comprehensive assessment stage begins. First, risk factor weights are configured; the weighted calculation logic is detailed below. Figure 5 The basic weights of risk factors are determined based on correlation analysis of 1200 historical incidents, with environmental risk at 0.4, traffic risk at 0.3, ship risk at 0.2, and geographical risk at 0.1, for a total weight of 1. The weights can be dynamically adjusted according to the navigation scenario. The adjustment formula is: the adjusted weight equals the basic weight multiplied by 1 plus the product of the adjustment coefficient and the scenario parameter. After adjustment, all factor weights are normalized to ensure the total remains 1. Typical adjustment parameters include: during the typhoon season (July-September), the environmental risk adjustment coefficient is 0.5, and the scenario parameter is typhoon warning levels 1 to 4; when oil tankers and chemical tankers are navigating, the ship risk adjustment coefficient is 1.0, and the scenario parameter is the IMDG cargo hazard class divided by 5; within port and waterway areas, the traffic risk adjustment coefficient is 0.8, and the scenario parameter is 1. The priority of weight adjustment decreases in the following order: extreme weather, ship type, navigation area, and season.

[0033] After the weights are configured, the risk monitoring range is defined as a 50-nautical-mile radius centered on the ship's current location. The basic risk value for each spatiotemporal grid within this range is calculated. The basic risk value is the sum of the products of the four risk characteristics and their corresponding weights, multiplied by 100, and converted into a score ranging from 0 to 100.

[0034] The basic risk value is then adjusted to obtain the actual navigation risk value. See the adjustment logic below. Figure 6 The revised calculation introduces data reliability and a ship risk resistance coefficient. The actual risk value equals the base risk value multiplied by the data reliability and then divided by the ship risk resistance coefficient. The data reliability value ranges from 0 to 1, determined by the overall data quality within the region. The ship risk resistance coefficient ranges from 0.35 to 1.0, calculated by weighting three indicators: wind resistance level, ship stability, and maneuverability, with weights of 0.4, 0.3, and 0.3 respectively. The base value is 0.35, and the weighted result of the performance indicators is multiplied by 0.65 and added to the base value to obtain the final coefficient. The wind resistance level coefficient is obtained by linear interpolation based on wind resistance level 5 (0.2) and level 12 (1.0). The ship stability coefficient is calculated according to the IMO stability standard, and the maneuverability coefficient decreases with increasing ship tonnage.

[0035] After completing the real-time risk calculation, the ship's future navigation path is predicted based on the inertial trajectory corrected by the channel boundary. The peak risk values ​​on the path in the next 1 hour, 3 hours and 6 hours are calculated respectively, and then weighted by weights of 0.5, 0.3 and 0.2 to obtain the comprehensive predicted risk value, so as to realize the early perception of risk.

[0036] After the risk assessment is completed, the process enters the early warning triggering and tiered response phase. See below for the early warning triggering logic. Figure 7 The system sets five core risk trigger conditions, and the early warning process is activated when any one of them is met: The first condition is that the comprehensive risk value of the ship's navigation area exceeds the medium risk threshold by 30 points within the next hour; the second condition is that the typhoon's path deviation exceeds the preset threshold, and the typhoon's 7-level wind circle will cover the ship's navigation area within 6 hours; the third condition is that there are no fewer than 5 ships within 3 nautical miles of the ship, and any two ships meet the condition that the nearest encounter distance is no more than 0.5 nautical miles and the nearest encounter time is no more than 10 minutes; the fourth condition is that the ship's own main engine, steering gear, or navigation equipment malfunctions, and the corresponding ship's own risk value, plus 20 points, exceeds the medium risk threshold; the fifth condition is that the ship enters a historically high-incidence area of ​​accidents, and the cosine similarity between the current sea state and the sea state at the time of the accident is no less than 0.7, with the similarity calculation parameters being the standardized values ​​of wind speed, significant wave height, and visibility.

[0037] After an alert is triggered, it is issued in a tiered manner based on the comprehensive risk value. The alert strategies for different levels are shown in Table 4. There are four risk levels: Low risk corresponds to 0 to 30 points, the system operates normally, the risk status is updated every 30 minutes, and no alert is issued; Medium risk corresponds to 31 to 60 points, a yellow alert is issued, text prompts are provided through the shipborne terminal, and risk details and precautions are pushed out every 10 minutes; High risk corresponds to 61 to 80 points, an orange alert is issued, the shipborne terminal activates an audible and visual alarm, risk details and preliminary handling suggestions are pushed out every 5 minutes, and the shore-based dispatch center is automatically notified; Extremely high risk corresponds to 81 to 100 points, a red alert is issued, the shipborne terminal activates an emergency audible and visual alarm, continuously pushes risk details and emergency handling suggestions, and automatically notifies the maritime regulatory authorities and the maritime search and rescue center.

[0038] Table 4. Early Warning Strategies for Different Levels

[0039] Based on an industry-standard navigation safety rule base and emergency response plan, the system generates differentiated and actionable response recommendations by considering risk level, risk type, and vessel and crew attributes. Typical response recommendations for different risk levels are shown in Table 5. Under medium-risk conditions, it is recommended that crew members strengthen lookout, adjust speed, and closely monitor changes in risk. Under high-risk conditions, it is recommended that crew members adjust course to avoid high-risk areas, reduce speed, and maintain a safe distance from nearby vessels. Under extremely high-risk conditions, it is recommended that crew members immediately take emergency avoidance measures, activate the emergency response plan, report to the maritime authorities, and prepare to abandon ship if necessary. The response recommendations are adapted to vessel type, tonnage, wind resistance, cargo hazard level, and crew experience to ensure their feasibility.

[0040] Table 5. Recommendations for handling different risk levels

[0041] The system utilizes a multi-terminal information synchronization link built through 5G and satellite communication to simultaneously push alarm information, risk details, and handling suggestions to shipboard navigation terminals, crew smart wristbands, and shore-based dispatch platforms, achieving real-time synchronization across multiple terminals. The ship-shore emergency communication channel possesses voice, video, and data transmission capabilities, facilitating remote guidance and support from shore-based dispatch personnel and maritime authorities in emergency situations. All early warning information and processing data are synchronized and stored at the shore-based control center, providing a basis for post-incident investigations and liability determination.

[0042] The working principle of this invention lies in mapping multi-source heterogeneous navigation data to the same spatiotemporal unit through a unified spatiotemporal index, and achieving deep weighted fusion by combining data credibility classification, thus solving the problems of shallow data splicing and failure to consider the reliability differences of data sources in traditional methods; achieving personalized risk assessment for each ship through multi-dimensional dynamic weights and ship risk resistance capability correction, thus solving the problems of single assessment dimensions and homogeneous standards in traditional methods; and achieving early risk perception and graded response through trajectory prediction and graded triggering mechanisms, combined with differentiated handling suggestions and multi-terminal collaboration, improving the effectiveness of early warning and the efficiency of handling, ultimately achieving a comprehensive improvement in the level of maritime navigation safety.

[0043] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for intelligent early warning of maritime vessel navigation risk classification based on multi-source spatiotemporal data fusion, characterized in that, Includes the following steps: S1 synchronously collects six types of heterogeneous navigation raw data: satellite navigation, AIS ship dynamics, marine meteorology, tides and currents, electronic nautical charts, and port traffic density. All raw data are preprocessed and standardized to obtain standardized basic data. S2 constructs a unified spatiotemporal index, and based on a three-level data credibility weighted fusion mechanism, performs spatiotemporal correlation fusion on standardized basic data to extract four types of standardized risk feature vectors: environmental risk features, traffic risk features, ship risk features, and geographical risk features. S3 configures the basic weights of multi-dimensional risk factors, dynamically adjusts the basic weights in combination with navigation scenarios and ship attributes, and calculates the basic risk values ​​of each spatiotemporal grid within the ship monitoring range based on the adjusted weights. S4 introduces a ship risk resistance capability correction coefficient to correct the basic risk value, obtains the actual navigation risk value specific to a single ship, and predicts the peak risk for multiple future periods based on the inertial trajectory corrected by the channel boundary. S5 sets up multiple core risk trigger conditions, combines actual navigation risk values ​​with predicted peak risks to conduct graded early warning judgments, generates differentiated handling suggestions based on risk level, risk type and vessel attributes, and pushes them to multiple terminals simultaneously to achieve ship-shore collaborative handling.

2. The intelligent early warning method for maritime vessel navigation risk classification based on multi-source spatiotemporal data fusion as described in claim 1, characterized in that, The preprocessing and standardization processes in step S1 include timestamp synchronization, coordinate system unification, missing value handling, and outlier removal; among which, missing value handling uses inverse distance weighted interpolation based on spatiotemporal proximity for completion, and the interpolation calculation formula is: ； in, For interpolation point values, Given the point value, The spatiotemporal distance between the known point and the interpolation point is given by k=2, where k=2 is the power exponent. Outlier removal uses the 3σ criterion: for continuous data of the same type within the same spatiotemporal grid, if data point x satisfies |x If μ|>3σ, it is considered an outlier and removed, where μ is the data mean and σ is the data standard deviation.

3. The intelligent early warning method for maritime vessel navigation risk classification based on multi-source spatiotemporal data fusion as described in claim 1, characterized in that, The three-level data credibility weighted fusion mechanism divides data into three levels, A, B, and C, based on source reliability, update frequency, and completeness, with corresponding weights of 1.0, 0.85, and 0.70, respectively. Level A corresponds to official authoritative sources with an update frequency of no more than 1 hour and a data integrity rate of no less than 95%; Level B corresponds to commercial data sources with an update frequency of no more than 6 hours and a data integrity rate of no less than 80%; and Level C corresponds to third-party shared data with an update frequency of no more than 24 hours and a data integrity rate of no less than 60%.

4. The intelligent early warning method for maritime vessel navigation risk classification based on multi-source spatiotemporal data fusion according to claim 1, characterized in that, The unified spatiotemporal index uses a 1km×1km spatial grid and a 15-minute time window. During spatiotemporal correlation weighted fusion, the multi-source features within the same spatiotemporal grid are weighted and averaged according to the data credibility weights to output the unified grid risk basic features. The fusion calculation formula is the sum of the products of each dimension feature value and the corresponding credibility weight divided by the sum of the total weights.

5. The intelligent early warning method for maritime vessel navigation risk classification based on multi-source spatiotemporal data fusion according to claim 1, characterized in that, The four types of standardized risk feature vectors are calculated as follows: Environmental risk characteristics: =0.4× +0.3× +0.2× +0.1× ; Among them, wind speed characteristics =min(V / 30,1), where V is the wind speed in m / s; significant wave height characteristic. =min(H / 10,1), where H is the wave height in meters; visibility characteristic. =1-min(S / 10,1), where S is visibility in nautical miles; rainfall intensity characteristics =min(R / 50,1), where R is the hourly rainfall in mm; Traffic risk characteristics: =min(N / 10,1), where N is the number of ships within a 3-nautical-mile radius; The inherent risk characteristics of the vessel: =0.4×tonnage+0.3×draft+0.3×cargo; Among them, tonnage characteristics =min(T / 100000,1), where T is the ship's deadweight tons; draft characteristics =min(D / 15,1), where D is the ship's draft in meters; cargo hazard rating characteristics. =C / 5, where C is the IMDG hazard class of the cargo and its value ranges from 1 to 5; Geographic risk characteristics: =0.6× +0.4× ; Among them, water depth characteristics =1-min(WD / 20,1), where WD is the current water depth in meters; characteristics of historical accident high-incidence areas. =1, the value is 1 when it is located in a high-incidence area, and 0 otherwise.

6. The intelligent early warning method for maritime vessel navigation risk classification based on multi-source spatiotemporal data fusion according to claim 1, characterized in that, The basic weights of the risk factors are: environmental risk 0.4, traffic risk 0.3, ship's own risk 0.2, and geographical risk 0.1; the dynamic weight adjustment formula is as follows: = ×(1+ ×α), after adjustment, normalize all factor weights to ensure that the sum of weights is 1; in For the first i The base weights of risk factors To correspond to the adjustment coefficient, α These are parameters associated with the scene.

7. The intelligent early warning method for maritime vessel navigation risk classification based on multi-source spatiotemporal data fusion according to claim 1, characterized in that, The formula for calculating the basic risk value is as follows: =( × + × + × + × )×100； The formula for calculating the actual navigation risk value is as follows: = Where T is the data reliability and its value ranges from 0 to 1, and C is the ship's risk resistance coefficient and its value ranges from 0.35 to 1.0; The formula for calculating the ship's risk resistance coefficient is C = 0.35 + 0.65 × (0.4 × ... +0.3× +0.3× ),in This is the wind resistance rating coefficient. For the ship's stability coefficient, For ship handling performance coefficient; The formula for calculating the risk value of future trajectory prediction is as follows: =0.5× +0.3× +0.2× ,in , , These are the peak risk values ​​for the next 1 hour, 3 hours, and 6 hours, respectively.

8. The intelligent early warning method for maritime vessel navigation risk classification based on multi-source spatiotemporal data fusion according to claim 1, characterized in that, The multiple core risk triggering conditions include five categories, and the early warning process will be initiated if any one of them is met: Category 1: The comprehensive risk value of the vessel in the navigation area within the next hour exceeds the medium risk threshold by 30 points. The second category is when the typhoon's path deviates beyond the preset threshold, and the typhoon's 7-level wind circle will cover the shipping area within 6 hours. Category 3: There are no fewer than 5 vessels within 3 nautical miles of the vessel, and any two vessels meet the condition that the closest encounter distance is no more than 0.5 nautical miles and the closest encounter time is no more than 10 minutes. The fourth category involves malfunctions in the ship's main engine, steering gear, or navigation equipment, which, when combined with the ship's own risk value by 20 points, exceeds the medium risk threshold. The fifth category is vessels that have entered areas with a high incidence of historical accidents, and the cosine similarity between the current sea state and the sea state at the time of the accident is not less than 0.

7.

9. The intelligent early warning method for maritime vessel navigation risk classification based on multi-source spatiotemporal data fusion according to claim 1, characterized in that, The tiered early warning system classifies navigation risks into four levels: Low risk corresponds to a comprehensive risk score of 0 to 30, with no early warning and the risk situation updated every 30 minutes; Medium risk corresponds to a comprehensive risk score of 31 to 60, and a yellow alert is issued. Text prompts are given through the shipboard terminal, and risk details are pushed out every 10 minutes. High risk corresponds to a comprehensive risk value of 61 to 80 points, an orange alert is issued, the shipborne terminal activates an audible and visual alarm, pushes risk details and preliminary handling suggestions every 5 minutes, and automatically notifies the shore-based dispatch center; Extremely high risk corresponds to a comprehensive risk value of 81 to 100 points, triggering a red alert. The shipborne terminal will activate an emergency audible and visual alarm, continuously push risk details and emergency response suggestions, and automatically notify the maritime regulatory authorities and the maritime search and rescue center.

10. The intelligent early warning method for maritime vessel navigation risk classification based on multi-source spatiotemporal data fusion according to claim 1, characterized in that, The differentiated handling recommendations are generated based on the navigation safety rule base and emergency response plan, combined with the ship type, tonnage, wind resistance level, cargo hazard level and crew attributes; multi-terminal synchronous early warning is generated through 5G and satellite communication links, and the early warning information is pushed to the shipborne navigation terminal, crew smart wearable devices and shore-based dispatch platform to realize ship-shore emergency communication and data synchronous retention.

Images