Ship collision risk early warning method and system

Abstract

This invention relates to the field of maritime traffic safety supervision technology, and provides a method and system for early warning of ship collision risks. The early warning method includes: acquiring ship image data and ship automatic identification system (AIS) data; determining whether the working status of the AIS data is normal or abnormal; dynamically switching the trajectory generation mode according to the working status and generating a corresponding target trajectory, wherein in the normal state, a fused trajectory is generated based on ship image data and AIS data, and in the abnormal state, a fallback trajectory is generated based on image data. This invention, through a dynamic switching mechanism based on working status, ensures uninterrupted early warning services when data is abnormal, and simultaneously improves system reliability and early warning accuracy by combining core collision indicators and dynamic safety zones to determine risk levels and filter invalid alarms.

Description

Technical Field

[0001] This application relates to the field of maritime traffic safety supervision technology, and in particular to a method and system for early warning of ship collision risks. Background Technology

[0002] Currently, in the field of maritime traffic safety supervision, ship collision risk warning mainly relies on data from the Automatic Identification System (AIS) or video surveillance data.

[0003] Existing technologies generally adopt early warning schemes that rely on a single AIS data source for independent analysis. They depend solely on AIS to collect dynamic information such as ship position, speed, and heading, and combine it with a fixed threshold model to analyze the ship encounter situation and determine the collision risk.

[0004] However, when the ship's AIS equipment is turned off, signal transmission is delayed, or data is distorted, the system often cannot obtain effective data, resulting in the interruption of the early warning function and making it difficult to ensure business continuity. Summary of the Invention

[0005] Therefore, it is necessary to provide a ship collision risk early warning method and system to address the above-mentioned technical problems. The system dynamically switches the trajectory generation mode based on the working status of the Automatic Identification System (AIS) data, automatically generates a fallback trajectory to ensure uninterrupted early warning when data is abnormal, and combines core collision indicators with dynamic safety zones to determine the risk level, thereby improving the reliability and accuracy of the system.

[0006] The first aspect of this invention provides a method for early warning of ship collision risks. The method includes the following steps: acquiring ship image data and ship automatic identification system (AIS) data; determining whether the working status of the AIS is abnormal based on the AIS data; dynamically switching the trajectory generation mode according to the working status and generating a corresponding target trajectory; wherein, when the AIS data is in a normal state, a fused trajectory is generated based on the ship image data and the AIS data as the target trajectory; when the AIS data is in an abnormal state, a fallback trajectory is generated based on the ship image data as the target trajectory.

[0007] In one embodiment, when the Automatic Identification System (AIS) data is in a normal state, the timestamp of the AIS data is extracted and compared with the local clock to calculate the delay duration. Based on the delay duration, the historical dynamic data of the AIS data is linearly interpolated to generate the current dynamic data of the ship, and the fused track is determined based on the dynamic data of the ship.

[0008] In one embodiment, spatiotemporal correlation matching is performed on ship image data and ship automatic identification system data to determine a fused track.

[0009] In one embodiment, spatiotemporal correlation matching includes coarse matching screening and fine matching screening, in which ship image data and ship automatic identification system data are sequentially subjected to coarse matching screening and fine matching screening.

[0010] In one embodiment, coarse matching screening uses a rectangular area as the matching range to initially screen ship targets, while fine matching screening adjusts the elliptical matching area according to ship dynamic data to calculate a weighted distance value, and determines the target track based on the weighted distance value.

[0011] In one embodiment, when the Automatic Identification System (AIS) data is in an abnormal state, the AIS data exhibits at least one of data interruption, data jump, or data distortion.

[0012] In one embodiment, ship features are extracted based on ship image data, latitude and longitude mapping and dynamic deviation correction are performed on the ship features to generate ship dynamic data, and a bottom-line track is determined based on the ship dynamic data.

[0013] In one embodiment, the warning method further includes determining the collision risk based on the target trajectory and outputting a warning signal.

[0014] In one embodiment, collision risk determination and outputting a warning signal includes the following steps: calculating core collision indicators between ships based on the target trajectory, constructing a safe zone based on at least one of the data from the Automatic Identification System (AIS) data or ship image data, determining the risk level based on the core collision indicators and the safe zone, filtering invalid alarms based on a preset electronic fence, and outputting a warning signal according to the risk level.

[0015] A second aspect of this invention provides a ship collision risk early warning system. The early warning system includes an information acquisition module, an anomaly determination module, a mode switching module, and a risk determination module. The information acquisition module acquires ship image data and ship automatic identification system (AIS) data. The anomaly determination module determines the data status of the AIS data. The mode switching module dynamically switches the trajectory generation mode based on the data status and generates a corresponding target trajectory. The risk determination module determines the collision risk based on the target trajectory and outputs an early warning signal. Specifically, when the AIS data is in a normal state, a fused trajectory is generated based on the ship image data and the AIS data as the target trajectory. When the AIS data is in an abnormal state, a fallback trajectory is generated based on the ship image data as the target trajectory.

[0016] The technical advantages of this invention are as follows: Through a dynamic mode switching mechanism for data status, when AIS data experiences interruptions, jumps, or distortions, it can switch to a pure video mode, using video positioning and trajectory prediction algorithms to generate target trajectories. This solves the problem of early warning service interruption caused by over-reliance on AIS data in existing technologies, thus improving system reliability. Simultaneously, by acquiring ship image data and AIS data and performing time synchronization and spatiotemporal correlation matching, this invention achieves precise alignment between video trajectories and AIS data, resolving data misalignment issues. Furthermore, by combining dynamic safety zone construction and electronic fence filtering technologies, it reduces the false alarm and missed alarm rates of the early warning system. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the ship collision risk warning method according to an embodiment of the present invention.

[0018] Figure 2 This is a schematic diagram of the structure of a ship collision risk warning system according to an embodiment of the present invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application. Example 1:

[0020] like Figure 1 As shown, this embodiment provides a ship collision risk early warning method, which solves the problem of early warning service interruption caused by over-reliance on ship automatic identification system data in the prior art. Through a dynamic switching mechanism, the accuracy of the trajectory is improved by fusing two data when the data is normal, and the early warning is not interrupted by generating a backup trajectory using image data when the data is abnormal, thus achieving all-weather, highly reliable maritime safety supervision.

[0021] Specifically, the method of this embodiment includes the following steps: Step S101: Acquire ship image data and ship automatic identification system data.

[0022] Specifically, ship image data can be acquired through video capture equipment deployed on shore-based infrastructure, such as bridges, lighthouses, and high-rise buildings on the shore. This equipment can be a high-definition network camera or an industrial camera, used to capture and monitor video streams of ships in the waters in real time. Automatic Identification System (AIS) data is acquired through an AIS receiver, which receives AIS messages sent by ships, containing dynamic information such as the ship's unique identifier, location, speed, and heading.

[0023] Step S102: Determine whether the working status of the Automatic Identification System (AIS) data is normal or abnormal.

[0024] In actual operation, AIS data may become abnormal due to signal obstruction, equipment failure, or other reasons. Abnormal states of AIS data specifically include one or more of the following: data interruption, data jumps, and data distortion. When AIS data is in a normal state, a fused track is generated by combining ship image data and AIS data, and this fused track is used as the target track. When AIS data is in an abnormal state, the system no longer relies on the failed AIS data, but only on the ship image data to generate a fallback track, which is used as the target track. This embodiment avoids early warning interruptions caused by the failure of a single AIS data source by dynamically switching data status determination and track generation modes, ensuring the continuous operation of maritime early warning services.

[0025] Step S103: Dynamically switch the track generation mode according to the working status and generate the corresponding target track.

[0026] Specifically, this is the core step of this embodiment. In actual operation, AIS data may become abnormal due to signal obstruction, equipment failure, or other reasons. In such cases, relying solely on AIS data will not be able to generate a track. This step introduces a dynamic switching mechanism to flexibly adjust the target track generation strategy based on the data status of the AIS data. When the AIS data is normal and valid, a fusion mode is used, combining video and AIS data to generate a high-precision track. When an abnormal AIS data status is detected, it automatically switches to a pure video mode, using only ship image data to calculate the ship's position and heading to generate the target track. This dynamic switching mechanism ensures that the system can still output continuous target tracks even if any single data source is missing, improving continuity.

[0027] Step S104: Configure the system to determine collision risk based on the target trajectory and output a warning signal; After obtaining the target trajectory, the system performs calculations based on a preset risk assessment logic. For example, it can calculate the closest encounter distance (DCPA) and closest encounter time (TCPA) between ships based on the target trajectory, or calculate the distance between the ship and fixed obstacles such as shore-based facilities and bridge piers. When the calculation results meet a preset risk threshold, the system determines that there is a collision risk and outputs a warning signal through audible and visual alarms, monitoring platform pop-ups, or SMS push notifications to remind supervisors or crew members to take collision avoidance measures. Through the above steps, this embodiment realizes a closed-loop process from data acquisition and fusion processing to risk warning, solving the problems of warning interruption and insufficient accuracy in the prior art. The system calculates core collision indicators between ships based on the target trajectory, constructs a safe zone based on at least one data point from the Automatic Identification System (AIS) data or ship image data, determines the risk level based on the core collision indicators and the safe zone, filters invalid alarms based on a preset electronic fence, and outputs a warning signal according to the risk level. Example 2:

[0028] Based on Embodiment 1 above, the preprocessing process of the Automatic Identification System (AIS) data is described in detail. In actual data transmission, there is a certain transmission delay between the ship's transmitter and the shore-based receiver. If the position information in the received message is directly used as the current position, it will cause a positional deviation for a high-speed moving ship. Therefore, after acquiring the AIS data, this embodiment further includes: extracting the data timestamp and comparing it with the local clock to calculate the delay duration; performing interpolation compensation based on historical data to obtain the ship parameters at the current moment.

[0029] Specifically, this preprocessing process can be performed after data acquisition in step S101 and before spatiotemporal correlation matching in step S201, and includes the following sub-steps: Step S201: Extract the timestamp and calculate the delay duration.

[0030] The system parses the received AIS message, extracts the original UTC timestamp information carried in the message, and reads the system's local high-precision clock time. By calculating the difference between the local received time and the original timestamp, the actual transmission delay of the message is obtained. For example, if the message shows a sending time of 10:00:00, while the local received time is 10:00:02, the delay is 2 seconds.

[0031] Step S202: Perform linear interpolation compensation on the historical dynamic data of the Automatic Identification System (AIS) based on the delay duration.

[0032] After determining the delay duration, the system needs to calculate the ship's true position at the current moment. This embodiment uses linear interpolation for compensation. The system retrieves historical AIS data of the ship prior to the current moment, such as the position points of the previous three moments, and constructs a position-time series. Based on the ship's average speed and heading during the historical period, combined with the calculated delay duration, the system calculates the ship's displacement increment during the delay period and superimposes this increment onto the original position in the message, thereby generating the ship's dynamic data at the current moment.

[0033] Step S203: Determine the target trajectory based on ship dynamic data.

[0034] It should be understood that linear interpolation assumes the ship moves at a constant speed in a straight line for a short period. However, for high-speed ships, such as those with speeds greater than 15 knots, if a turn occurs during the delay period, simple linear interpolation will cause the calculated position to deviate from the actual trajectory. Therefore, this embodiment introduces a heading correction coefficient. When the ship's speed exceeds a preset threshold and the rate of change of heading exceeds a preset range, the system corrects the interpolated trajectory by arc based on the heading change trend. The correction coefficient ranges from, for example, 0.5 to 1.0, and the specific value can be dynamically adjusted according to the actual turning angle to fit the circular trajectory of the ship during turning, thereby ensuring compensation accuracy in high-speed turning scenarios.

[0035] Through the aforementioned delay compensation and interpolation processing, this embodiment eliminates the positional error caused by the lag in AIS data transmission, making the data input to the spatiotemporal correlation matching module more timely and accurate, thus laying a solid data foundation for subsequent accurate matching and risk assessment. Example 3:

[0036] Based on Example 2, this embodiment addresses the issue that in actual water monitoring environments, there may be dozens or even hundreds of AIS targets simultaneously. Directly calculating the distance between each pixel or detection box in the video image and all AIS targets with high precision would impose a huge computational burden and make it difficult to meet the needs of real-time early warning. Therefore, this embodiment performs spatiotemporal correlation matching between ship image data and ship automatic identification system data to balance computational efficiency and matching accuracy.

[0037] Specifically, spatiotemporal correlation matching includes coarse matching and fine matching, which can be broken down into the following steps: Step S301: Perform coarse matching and filtering.

[0038] In this embodiment, coarse matching is target filtering based on rectangular regions. Specifically, the system first delineates an initial geofence rectangular region based on the field of view of the video surveillance image. The length and width of this rectangular region can be determined according to the actual waterway width. For example, in inland waterways, it can be set as a 50-meter × 30-meter rectangle. The system compares the positions of all vessels in the AIS data list with this rectangular region, retaining only potential targets located within the rectangular region. Rectangular regions have the advantage of simple calculation; irrelevant AIS targets outside the field of view can be quickly eliminated by comparing latitude and longitude coordinates, thereby reducing the amount of data processing required for subsequent fine matching.

[0039] As an alternative implementation, the size of the rectangular area can also be dynamically adjusted according to the busyness of the monitored waters. For example, in port areas with high ship density, the rectangular area can be appropriately reduced to further reduce the computational load.

[0040] Step S302: Perform fine matching and alignment.

[0041] After coarse matching and screening, the number of remaining potential targets has been reduced. At this point, high-precision fine matching is performed. In this embodiment, fine matching is based on dynamically adjusting the elliptical matching area according to the ship's speed and heading, and aligning it by combining the weighted distance of geographical distance and heading deviation. Specifically, the construction of the elliptical matching region takes into account the ship's motion characteristics. During navigation, the ship's position uncertainty mainly extends along the course, while the uncertainty in the direction perpendicular to the course is smaller. Therefore, using an ellipse shape better matches the ship's motion trajectory distribution than a circle or rectangle. The major axis L and minor axis W of the ellipse are not fixed values, but rather have a functional relationship with the ship's motion state. For example, the formula for calculating the major axis L is L = kv × v + L0, where v is the ship's current speed, kv is the speed influence coefficient (e.g., ranging from 0.8 to 1.2), and L0 is the basic major axis length (e.g., 10 meters). This formula reflects the influence of speed on position prediction uncertainty: the faster the speed, the greater the range of position change per unit time, therefore the search range (major axis) should be extended accordingly. Similarly, the minor axis W can also be adjusted based on the ship's width or course stability. Through this dynamic adjustment mechanism, the system can adapt to the ship matching needs under different navigation states, reducing missed matches caused by an excessively small or large search range.

[0042] After determining the elliptical matching region, the system further calculates the weighted distance between the video-detected target and the potential AIS target. The calculation of the weighted distance D takes into account both geographical distance and heading deviation. The calculation formula is, for example, D = w1 × D geo +w2×D heading D geo To map the geographical distance between the latitude and longitude of the video detection box center and the latitude and longitude of the AIS target, Dheading To normalize the angular deviation between the video detection heading and the AIS heading, w1 and w2 are weighting coefficients, for example, w1 is set to 0.7 and w2 to 0.3. The heading deviation is introduced as a weighting factor because heading is a crucial characteristic of ship motion; if two targets are geographically close but have significantly different headings, the probability of them matching is extremely low. By calculating the weighted distance, the system can filter out the true target with the most consistent motion state from multiple candidate targets that are spatially close, thereby improving the accuracy of association matching.

[0043] This embodiment employs a hierarchical strategy of rectangular coarse matching and elliptical fine matching. First, it uses rectangular screening to quickly eliminate irrelevant targets, and then uses an elliptical model that fits the motion characteristics for precise alignment. This ensures both the real-time response speed of the system and the high accuracy of multi-source data fusion, thus solving the problem of track misalignment in high-speed motion scenarios. Example 4:

[0044] Based on Example 3, this embodiment provides a detailed explanation of the specific implementation process of the mode switching step. In actual navigation environments, AIS signals are easily affected by terrain obstruction, electromagnetic interference, or equipment failure, leading to data anomalies. If the system relies solely on AIS data for track generation, the early warning function will immediately fail once the data is abnormal. This embodiment ensures the continuity of early warning services through anomaly detection and mode switching mechanisms.

[0045] Specifically, the data status includes at least one of data interruption, data jump, or data distortion. When the Automatic Identification System (AIS) data is determined to be abnormal, the system switches to pure video mode to generate a fallback track as the target track.

[0046] Step S401: Determine if AIS data is abnormal.

[0047] The system monitors the AIS data stream in real time and determines it to be abnormal when any of the following situations occur: First, data interruption, that is, no AIS messages are received within a preset interruption threshold time, such as 10 seconds, indicating that the signal link may be broken; Second, data jump, that is, the distance difference between the currently received position data and the position data at the previous moment exceeds the physical movement limit of the ship at the corresponding speed. For example, at a speed of 20 knots, the distance moved in 1 second cannot exceed 10 meters, indicating that there is noise or error in the data; Third, data distortion, that is, the message check bit is incorrect or the key field, such as the MMSI code format is invalid.

[0048] Step S402: Switch to pure video mode and generate target track.

[0049] Once the system detects any of the above situations, it switches to pure video mode, extracts ship features based on ship image data, generates ship positioning data by combining latitude and longitude mapping and dynamic deviation correction, and predicts future tracks through track prediction algorithms to generate target tracks.

[0050] Specifically, the pure video mode processing logic comprises two key stages: First, the video localization stage. The system utilizes image processing algorithms, such as convolutional neural networks, to extract ship features from video frames, including hull outline and heading. Subsequently, combined with a pre-calibrated latitude and longitude mapping model, the image pixel coordinates are converted into geographic latitude and longitude. Considering that shore-based cameras may experience slight vibrations due to wind and temperature, the system also introduces a dynamic deviation correction mechanism. By analyzing the displacement of background feature points, the mapping parameters are corrected in real time to ensure that the positioning accuracy is not less than 90% of that under normal AIS conditions. Second, the trajectory prediction stage. Since video localization only provides the current position and cannot directly obtain speed and heading information, the system uses trajectory prediction algorithms, such as the Kalman filter algorithm, to estimate the ship's motion state. The algorithm uses historical video localization data to construct a motion model and predict the ship's position in the short future, such as 3 to 5 seconds, thereby generating a continuous target trajectory. Kalman filtering is only a preferred implementation method in this embodiment; other prediction algorithms, such as particle filtering and linear regression, are also within the scope of protection of this invention.

[0051] This embodiment utilizes the above mechanism to fill the data gaps in the extreme case of AIS data failure, thereby achieving uninterrupted early warning service continuity. This dynamic switching and predictive compensation design solves the system vulnerability problem caused by the over-reliance on a single data source in existing technologies. Example 5:

[0052] Based on Example 4, this embodiment provides a detailed explanation of the specific implementation process of collision risk determination. In the prior art, a fixed distance threshold is often used, such as uniformly setting 50 meters as the safe distance for risk determination. However, ships of different sizes, such as cargo ships of thousands of tons and small fishing boats, have large differences in inertia, and the braking distance at different speeds is also very different. A fixed threshold often leads to a lag in early warning when large ships are sailing at high speeds. Therefore, this embodiment introduces a dynamic safe zone construction mechanism and an invalid alarm filtering mechanism.

[0053] Specifically, collision risk assessment includes the following steps: Step S501: Calculate the core collision indicators between ships based on the target trajectory.

[0054] Based on the aforementioned target track, which is either a fused track under normal AIS conditions or a video fallback track under abnormal AIS conditions, the motion parameters such as the position, speed, and heading of each vessel are calculated in real time. Then, core collision indicators such as the encounter distance, relative speed, heading intersection angle, closest encounter distance (CPA), and closest encounter time (TCPA) are calculated as the quantitative basis for subsequent risk level assessment.

[0055] Step S502: Construct a safe zone based on at least one of the data from the Automatic Identification System (AIS) data and the ship image data.

[0056] Based on Automatic Identification System (AIS) data and / or ship image data, a personalized dynamic safety zone is independently constructed for each ship. When AIS data is in a normal state, ship attributes such as length and beam, as well as real-time speed, are directly read from AIS messages. When AIS data is abnormally interrupted, jumps, or distorted, the ship dimensions are inferred from the ship image data through target detection, feature extraction, and coordinate mapping, and the real-time speed is calculated using the temporal changes of multiple video frames.

[0057] Step S503: Determine the risk level by combining the core collision indicators with the dynamic safety zone.

[0058] The calculated core collision indicators are spatially and situationally coupled and compared with the ship's own dynamic safety zone. By combining the nearest encounter distance, encounter time and the relative motion of the two ships, different collision risk levels are classified to form a standardized risk classification result, which provides a basis for differentiated early warning output.

[0059] Step S504: Filter invalid alarms through preset electronic fence.

[0060] The actual monitored waters include the main channel and branch channels, as well as non-effective navigation areas such as shoals, mudflats, enclosed waters, construction restricted areas, and berthing areas. When vessels are normally moored, berthed, or operating in these areas, there is no actual risk of collision. Triggering warnings as usual would increase the workload of monitoring personnel.

[0061] This system pre-defines electronic fences using latitude and longitude coordinates to define the boundaries of effective navigation areas and restricted navigation areas. After completing the risk level assessment and before officially issuing an early warning, it verifies whether the target vessel's current position is within the effective navigation area defined by the electronic fence. If it is in a non-navigable area, it is determined as an invalid alarm and directly filtered, without entering the early warning push process; alarm logic is only retained for high-risk encounter targets within the effective navigation area.

[0062] Step S505: Output the corresponding early warning signal according to the risk level.

[0063] Based on the classified collision risk levels, different levels of early warning signals are output to achieve hierarchical management. Low-risk situations are only recorded in the background; medium-risk situations are indicated by pop-up windows on the interface and the situation is highlighted in red; high-risk situations are simultaneously triggered by platform alarms, which intuitively present the vessel's position, encounter situation and risk level, making it easier for regulatory personnel to carry out timely dispatch and intervention. Example 6:

[0064] This embodiment provides a ship collision risk early warning system. This system and the method embodiment belong to the same inventive concept.

[0065] like Figure 2 As shown, the system includes an information acquisition module 601, an anomaly detection module 602, a mode switching module 603, and a risk detection module 604.

[0066] The information acquisition module 601 is configured to acquire ship image data and ship automatic identification system (AIS) data, and to perform time synchronization processing on the various data sources. Specifically, the data acquisition and synchronization module is the perception layer of the system, and its hardware carrier can include devices such as high-definition network cameras, AIS signal receiving antennas, and decoders deployed on shore. This module is not only responsible for acquiring raw data, but also has a built-in clock synchronization unit. By receiving the timestamps from the Network Time Protocol (NTP) server, it uniformly calibrates the internal clocks of the camera and the AIS receiver, ensuring the consistency of video streams and AIS messages in the time dimension. The data acquisition and synchronization module can be an integrated industrial control computer or a distributed combination of acquisition terminals, as long as it can achieve synchronous acquisition of multi-source data.

[0067] The anomaly determination module 602 is configured to determine whether the working status of the Automatic Identification System (AIS) data is abnormal. The working status includes normal status or abnormal status. This module is a key control unit to ensure the continuity of system business. The key control unit continuously monitors the data status of AIS data. Once an anomaly such as data interruption, jump or distortion is detected, the mode switching logic is triggered, and the positioning and prediction algorithm in pure video mode is called to generate a fallback track as the target track.

[0068] The mode switching module 603 is configured to dynamically switch the track generation mode according to the working status and generate the corresponding target track. When the Automatic Identification System (AIS) data is in a normal state, a fused track is generated based on the ship image data and AIS data as the target track; when the AIS data is in an abnormal state, a fallback track is generated based on the ship image data as the target track.

[0069] The risk assessment module 604 is configured to assess collision risk based on the target trajectory. This module receives the target trajectory output by the mode switching and trajectory generation module and performs calculations based on preset risk assessment logic. Specifically, the risk assessment module internally stores electronic chart data and electronic fence data. First, based on at least one of the Automatic Identification System (AIS) data and ship image data, it determines the ship's real-time dimensions and speed, dynamically constructing a safe zone. Then, it calculates core collision indicators between ships based on the target trajectory, including the closest encounter distance (DCPA) and the closest encounter time (TCPA). Finally, it determines the risk level based on the core collision indicators and the safe zone, filters invalid alarms through a preset electronic fence, outputs a warning signal according to the risk level, and drives peripheral devices to respond. For example, it can control an audible and visual alarm via a relay for on-site warnings, or push warning SMS messages or pop-up messages to the maritime regulatory terminal via a mobile communication network. Example 7:

[0070] To more intuitively demonstrate the practical application effect of the technical solution of the present invention, this embodiment takes a complex waterway monitoring scenario of an inland waterway as an example to describe in detail the complete operation process of the ship collision risk early warning method and system. The specific parameter values ​​in this embodiment are only used to explain the present invention and do not constitute a limitation on the scope of protection of the present invention.

[0071] In this application scenario, the monitored waterway is an inland waterway approximately 300 meters wide, with high-definition network cameras deployed on both banks as video acquisition devices, and AIS receiving base stations also provided. All devices in the system use an NTP server for unified time synchronization, with a preset threshold of ±50ms for time synchronization processing to ensure consistent time references across all data sources.

[0072] The scenario process is described as follows: Vessel A is a cargo ship with a length of 80 meters, which entered the monitored waters normally at a speed of 12 knots.

[0073] Phase 1: Data Acquisition and Normal Integration.

[0074] The system first acquires ship image data and Automatic Identification System (AIS) data. The video acquisition device captures the image stream of ship A in real time, and the AIS receiving base station receives the AIS message sent by ship A. Due to a delay of approximately 2 seconds in AIS message transmission, the system immediately executes delay compensation logic after acquiring the data: extracting the data timestamp and comparing it with the local clock to calculate the delay duration, and performing interpolation compensation based on historical data to obtain the ship parameters at the current moment. Specifically, the system uses linear interpolation, combined with ship A's historical speed and heading, to calculate the precise latitude and longitude position at the current moment. Subsequently, the system performs spatiotemporal correlation matching on the video data and the compensated AIS data. The matching process adopts a hierarchical strategy: first, a coarse matching filter is performed, setting the rectangular area to 50m × 30m to quickly eliminate irrelevant targets outside the field of view; then, a fine matching alignment is performed, dynamically adjusting the elliptical matching area based on the ship's speed and heading, and aligning it using a weighted distance calculated from geographical distance and heading deviation. Since the data source is normal, the system outputs the target trajectory in a fusion mode based on the correlation matching results and determines the collision risk based on the target trajectory. At this point, the system dynamically constructs a safe zone based on the ship's dimensions and speed. The extension radius R is calculated using the formula R=k×v+R0, and the sector angle is set to 75° to cover the risk zone in the bow direction.

[0075] Phase Two: Anomaly Triggering and Mode Switching. When vessel A reaches a bend in the channel, its AIS signal is suddenly interrupted due to terrain obstruction, and the interruption lasts for more than a preset threshold (e.g., 10 seconds). The system detects that the vessel's AIS data is interrupted and determines that the data is abnormal. At this point, the system immediately triggers a dynamic switching mechanism, automatically switching to pure video mode to generate the target track. In pure video mode, the system extracts the contour features of vessel A based on the vessel image data, combines latitude and longitude mapping with dynamic deviation correction to generate vessel positioning data, and predicts the future track using a Kalman filter algorithm to generate the target track. The prediction duration is set to 3 to 5 seconds to ensure that the system can still output continuous track information during the gap period of missing AIS data.

[0076] Phase 3: Risk Warning and Business Continuity. During pure video mode operation, the system continuously assesses collision risk based on the target track. At this time, vessel A and vessel B, moored on the side of the channel, gradually approach each other. The system determines that there is a collision risk based on vessel A's dynamic safety zone, calculating that the closest encounter distance (DCPA) is less than the safety threshold. Simultaneously, the system combines electronic fence filtering to remove invalid alarms from non-navigable areas, confirming that vessel B is within the navigable area and the alarm is valid. The system then outputs a warning signal, prompting supervisors to pay attention to collision avoidance through a pop-up window on the monitoring platform. Supervisors intervene in a timely manner to guide vessel A, which successfully avoids the collision. Approximately 30 seconds later, vessel A leaves the obstructed area, the AIS signal returns to normal, and the system automatically switches back to fusion mode to continue providing high-precision track services.

[0077] As can be seen from the above application scenarios, the ship collision risk early warning method and system provided in this embodiment can ensure uninterrupted operation of the early warning service through dynamic switching and fallback trajectory generation mechanisms in extreme situations such as data interruption of the Automatic Identification System (AIS). Meanwhile, the hierarchical matching strategy and dynamic safety zone construction mechanism improve the accuracy of early warnings in complex water environments, verifying the practical feasibility and high reliability of the technical solution.

[0078] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0079] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for early warning of ship collision risks, characterized in that, The method includes the following steps: Acquire ship image data and ship automatic identification system data; Based on the data from the Automatic Identification System (AIS), determine whether its operational status is abnormal; and The trajectory generation mode is dynamically switched according to the working status, and the corresponding target trajectory is generated. Specifically, when the Automatic Identification System (AIS) data is in a normal state, a fused track is generated based on the ship image data and the AIS data as the target track; when the AIS data is in an abnormal state, a fallback track is generated based on the ship image data as the target track.

2. The ship collision risk early warning method according to claim 1, characterized in that, When the Automatic Identification System (AIS) data is in a normal state, the timestamp of the AIS data is extracted and compared with the local clock to calculate the delay duration. Based on the delay duration, the historical dynamic data of the AIS data is linearly interpolated to generate the current dynamic data of the ship, and the fused track is determined based on the dynamic data of the ship.

3. The ship collision risk early warning method according to claim 2, characterized in that, The ship image data and the ship automatic identification system data are subjected to spatiotemporal correlation matching to determine the fused track.

4. The ship collision risk early warning method according to claim 3, characterized in that, The spatiotemporal correlation matching includes coarse matching and fine matching, and the ship image data and the ship automatic identification system data are subjected to coarse matching and fine matching in sequence.

5. The ship collision risk early warning method according to claim 4, characterized in that, The coarse matching screening uses a rectangular area as the matching range to initially screen ship targets. The fine matching screening adjusts the elliptical matching area according to the ship dynamic data, and calculates a weighted distance value based on geographical distance and heading deviation. The fused track is determined based on the weighted distance value.

6. The ship collision risk early warning method according to claim 1, characterized in that, When the Automatic Identification System (AIS) data is in an abnormal state, the AIS data is subject to at least one of the following: data interruption, data jump, or data distortion.

7. The ship collision risk early warning method according to claim 6, characterized in that, Ship features are extracted based on the ship image data, latitude and longitude mapping and dynamic deviation correction are performed on the ship features to generate ship dynamic data, and the bottom-line track is determined based on the ship dynamic data.

8. The ship collision risk early warning method according to claim 1, characterized in that, The early warning method also includes determining the collision risk based on the target trajectory and outputting an early warning signal.

9. The ship collision risk early warning method according to claim 8, characterized in that, The collision risk assessment and early warning signal output includes the following steps: calculating the core collision index between ships based on the target trajectory; constructing a safe zone based on at least one of the data from the Automatic Identification System (AIS) or the ship image data; determining the risk level based on the core collision index and the safe zone; filtering out invalid alarms based on a preset electronic fence; and outputting an early warning signal according to the risk level.

10. A ship collision risk early warning system, characterized in that, The early warning system includes: The information acquisition module acquires ship image data and data from the Automatic Identification System (AIS). The anomaly detection module determines whether the automatic identification system (AIS) is malfunctioning based on the data from the system. The mode switching module dynamically switches the trajectory generation mode based on the operating state and generates the corresponding target trajectory; and The risk assessment module assesses the collision risk based on the target trajectory and outputs a warning signal. Specifically, when the Automatic Identification System (AIS) data is in a normal state, a fused track is generated based on the ship image data and the AIS data as the target track; when the AIS data is in an abnormal state, a fallback track is generated based on the ship image data as the target track.

Images