Ship operation collision avoidance method and system

Abstract

The invention relates to the technical field of ship collision avoidance, and discloses a ship operation collision avoidance method and system, and the method comprises the steps: obtaining the absolute motion data of a target relative to a seabed, judging the absolute motion state (static or dynamic) of the target according to the data, and carrying out the qualitative processing of the target. According to the invention, a fixed navigation mark and a drifting fishing gear or fragments can be effectively distinguished, and an automatic control system is prevented from executing collision avoidance operation based on a moving object which is wrongly recognized, so that the risk of collision, stranding or other navigation accidents is remarkably reduced. According to the technical scheme, a more reliable and more intelligent environment perception and collision avoidance decision-making mechanism is provided for a ship navigation system, and the navigation safety and automation level of a ship in a complex marine environment are improved.

Description

Technical Field

[0001] This invention relates to the field of ship collision avoidance technology, and in particular to a ship operation collision avoidance method and system. Background Technology

[0002] In modern shipping, especially with highly automated or unmanned vessels, visual sensors are often used to identify navigational aids and avoid collision risks. Typically, optical cameras mounted on the bow of the vessel capture images of distant navigational aids. The system then analyzes these images, extracts features, and matches them against a pre-set navigational aid database to determine the type of the aid, estimate its distance and bearing, and thus guide the vessel safely along a predetermined route.

[0003] However, actual navigation environments are often complex and dynamic. For example, in coastal or inland waterways where fishing operations are frequent, visual systems often encounter a large number of non-standard floating objects, such as various fishing gear buoys and marine debris. These objects may resemble regular navigation marks in shape and color, leading to frequent misjudgments by existing recognition algorithms: either mistaking fishing gear for navigation marks, generating false alarms; or failing to identify genuine navigation marks in time due to visual interference, resulting in missed detections. The input of a large amount of misleading visual information significantly increases the uncertainty of the system's perception of navigable areas.

[0004] Furthermore, these types of fishing gear and other floating objects are usually not stationary, but rather drift with the current or change position due to the influence of fishing vessels. When the system misidentifies a moving fishing gear as a fixed navigation mark, it will observe an unwarranted change in the "marker's" position, causing system logic confusion. If the automatic control unit performs collision avoidance actions based on such erroneous identification results, it may instead lead the vessel to a potentially dangerous area.

[0005] To improve system robustness, modern ship navigation systems typically integrate information from multiple sensors, such as radar and AIS. However, in waters densely populated with fishing vessels, radar echoes often contain numerous small, unclassified targets, making it difficult to reliably distinguish between official navigation marks and fishing gear buoys of similar size; furthermore, AIS usually does not provide identification for passive navigation marks or fishing gear. This results in multiple sensors being able to collectively confirm the presence of an object, but struggling to accurately determine its true nature—especially when the target is visually highly similar to a navigation mark. This inconsistency between sensor information significantly reduces the system's confidence level in perceiving the surrounding environment.

[0006] When ships need to perform complex and time-sensitive maneuvers, the aforementioned persistent perceptual ambiguity and sensor conflicts become particularly critical. If the system frequently processes false alarms or struggles to reliably identify real navigational aids, its overall confidence in the accuracy of environmental modeling will significantly decrease. The automated control system thus faces a dilemma: should it accept ambiguous real-time sensing data or rely on static navigational chart information? This uncertainty can easily lead to decision-making delays, operational hesitation, and even the execution of erroneous automated commands, thereby significantly increasing the risk of navigational accidents such as collisions and groundings.

[0007] The existing solutions still have significant shortcomings in addressing the aforementioned technical bottlenecks and urgently need further optimization and improvement. Summary of the Invention

[0008] This invention provides a collision avoidance method and system for ship operations, aiming to solve the technical problems of existing ship navigation systems in complex marine environments, especially in areas with intensive fishing activities, such as fuzzy identification of dynamic obstacles, both navigational aids and non-navigational aids, frequent false alarms and missed alarms, decreased confidence of sensor data, and the resulting decision delays and increased operational risks.

[0009] Firstly, in order to solve the above-mentioned technical problems, the present invention provides a method for avoiding collisions during ship operations, comprising: Acquire the absolute motion data of the target relative to the seabed within a preset time period; Based on the absolute motion data, the absolute motion state of the target relative to the seabed is determined, and the absolute motion state includes a static state or a dynamic state. The target is qualitatively processed based on the absolute motion state; Based on the qualitative target information, collision avoidance control commands for the ship are generated; The qualitative processing includes: if the target is in the stationary state, the qualitative target information is used to initiate the navigation mark identification process; if the target is in the dynamic state, the qualitative target information is used to identify a non-navigation mark dynamic obstacle.

[0010] Preferably, the qualitative processing of the target based on the absolute motion state includes: Assess the intensity of local environmental disturbances and the stability of local water surface textures around the target; Assess the correlation between the target and navigational aids on electronic nautical charts; The confidence level of the target motion state is calculated based on the absolute motion data, the intensity of local environmental disturbance, the stability of local water surface texture, and the correlation degree of electronic nautical chart beacons. Based on the confidence level of the target's motion state, the target is qualitatively processed; The qualitative processing includes: when the confidence level of the target's motion state is higher than a first threshold, initiating the navigation mark identification process; when the confidence level of the target's motion state is lower than a second threshold, classifying the target as a non-navigation mark dynamic obstacle; and when the confidence level of the target's motion state is between the first threshold and the second threshold, initiating the enhanced observation process.

[0011] Preferably, when the confidence level of the target motion state is between the first threshold and the second threshold, the enhanced observation process is initiated, including: The adjustment parameters of the vessel are obtained, including the vessel speed, current channel width, importance level of the target on the electronic chart, and local environmental disturbance index. Adjust the first threshold and the second threshold according to the adjustment parameters; The confidence level of the target motion state is evaluated based on the adjusted first threshold and second threshold, and the enhanced observation process is initiated based on the evaluated confidence level.

[0012] Preferably, the step of calculating the target motion state confidence level based on the absolute motion data, the local environmental disturbance intensity, the local water surface texture stability, and the electronic chart beacon correlation degree includes: The corresponding weights are dynamically adjusted based on the reliability of the absolute motion data, the reliability of the local environmental disturbance intensity, the reliability of the local water surface texture stability, and the reliability of the electronic chart beacon correlation. The confidence level of the target motion state is calculated based on the adjusted weights.

[0013] Preferably, before dynamically adjusting the corresponding weights, the method further includes: The reliability of obtaining the absolute motion data; The reliability of acquiring the absolute motion data includes: Acquire the first estimate of the target motion data from the visual sensor; Acquire a second estimate of the target motion data from the radar system; Determine whether there is a conflict between the first estimate and the second estimate; When the conflict exists, the health status of the visual sensor and the radar system is obtained, and the corresponding weights are adjusted according to the health status. The extent to which current environmental parameters affect the visual sensor and the radar system is determined. Based on the degree of impact, the weights are adjusted again; The reliability of the absolute motion data is calculated based on the adjusted weights.

[0014] Preferably, the reliability of acquiring the absolute motion data includes: Acquire the first motion data of the target before occlusion, the first motion data including motion trend information, motion speed information, motion direction information, motion acceleration information and motion angular velocity information; Acquire second motion data of the target in the area where the shading occurs, the second motion data including local water flow direction information, local water flow velocity information, local wind direction information, and local wind speed information; Based on the first motion data and the second motion data, predict the possible range of motion of the target during the occlusion period; After the occlusion is removed, the third motion data of the target is acquired; Compare the third motion data with the possible motion range; The reliability of the absolute motion data is determined based on the comparison results.

[0015] Preferably, acquiring the first motion data of the target before occlusion includes: Continuously monitor the rate of change of the target's position, rate of change of velocity, and rate of change of direction; When the rate of change of the target's velocity or the rate of change of its direction exceeds a preset threshold, it is determined that the target is undergoing nonlinear motion. When the target is undergoing nonlinear motion, motion data from a short time window is used for analysis. Track the instantaneous angular acceleration of the target at the inflection point and make predictions around the inflection point.

[0016] Preferably, the instantaneous acceleration of the tracked target at the turning point includes: Get the current environment parameters; Based on the current environmental parameters, assess the extent to which the current environmental parameters affect the performance of the visual sensor and the radar system; Obtain the historical acceleration trend of the target before and after the inflection point; Compare the current instantaneous acceleration with the historical acceleration trend for consistency; Adjust the weight of instantaneous acceleration assessment based on the consistency comparison results; Based on the adjusted weights, the consistency comparison results, and the degree of influence, the instantaneous acceleration of the target at the inflection point is calculated.

[0017] Preferably, obtaining the current environmental parameters includes: Obtain the ship's own position information; Obtain the position information of the target relative to the ship; Based on the ship's own position information and the target's position information relative to the ship, calculate the absolute position of the target relative to the seabed; A local environmental perception area is defined with the absolute position of the target relative to the seabed as the center; The shipborne local environment sensor is activated to scan the local environment sensing area and obtain the visibility, precipitation intensity, sea clutter intensity, water flow direction, water flow velocity, wind direction, and wind speed within the local environment sensing area to generate local environmental parameters of the target area.

[0018] Secondly, the present invention provides a collision avoidance system for ship operations, comprising: The detection end is used to acquire the absolute motion data of the target relative to the seabed within a preset time period; The processing unit is used to determine the absolute motion state of the target relative to the seabed based on the absolute motion data, wherein the absolute motion state includes a static state or a dynamic state; and to perform qualitative processing on the target based on the absolute motion state. The control unit is used to generate collision avoidance control commands for the ship based on the qualitative target information. The qualitative processing includes: if the target is in the stationary state, the qualitative target information is used to initiate the navigation mark identification process; if the target is in the dynamic state, the qualitative target information is used to identify a non-navigation mark dynamic obstacle.

[0019] The collision avoidance method and system disclosed in this application acquires the absolute motion data of a target relative to the seabed and determines the target's absolute motion state (stationary or dynamic) based on this data, thereby qualitatively classifying the target. Specifically, if the target is stationary, a navigation mark identification process is initiated; if the target is dynamic, it is classified as a non-navigation mark dynamic obstacle. This method effectively solves the problem in existing technologies, especially in areas with intensive fishing activities, where visual recognition systems frequently misidentify non-standardized objects such as fishing gear as navigation marks, or where visual noise causes the failure to report genuine navigation marks. Through preliminary judgment of the target's motion state, this application can effectively distinguish between fixed navigation marks and drifting fishing gear or debris, avoiding the automated control system from performing collision avoidance operations based on incorrectly identified moving objects, thus significantly reducing the risk of collisions, groundings, or other navigation accidents. This technical solution provides a more reliable and intelligent environmental perception and collision avoidance decision-making mechanism for ship navigation systems, improving the navigation safety and automation level of ships in complex maritime environments. Attached Figure Description

[0020] Figure 1 This is a schematic flowchart of a collision avoidance method for ship operations provided in an embodiment of the present invention; Figure 2This is a schematic flowchart of another collision avoidance method for ship operations provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of a ship collision avoidance system provided in an embodiment of the present invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] Reference Figure 1 The present invention provides a schematic flowchart of a collision avoidance method for ships, comprising the following steps: S1, acquire the absolute motion data of the target relative to the seabed within a preset time period; S2, based on the absolute motion data, determine the absolute motion state of the target relative to the seabed, the absolute motion state including a static state or a dynamic state; S3, perform qualitative processing on the target based on the absolute motion state; S4, based on the qualitative target information, generates collision avoidance control commands for the ship; The qualitative processing includes: if the target is in the stationary state, the qualitative target information is to initiate the navigation mark identification process; if the target is in the dynamic state, the qualitative target information is to identify a non-navigation mark dynamic obstacle; the control command includes at least the command to control the current operation mode, the boundary of the operation area, and the key operation equipment.

[0023] To better understand the technical solution proposed in this application, some key terms involved will be explained first.

[0024] Among them, the "seabed" is suitable for both ocean-going vessels and inland waterway vessels.

[0025] "Target" refers to any maritime object that may affect the safety of a ship's navigation during navigation, including but not limited to navigation marks, fishing gear, floating objects, and other vessels. "Absolute motion data" refers to the target's motion information relative to a fixed reference frame (e.g., the seabed), typically including parameters such as position, velocity, and acceleration. "Preset time" refers to the time window set when acquiring and analyzing target motion data; the length of this time window can be adjusted according to the actual application scenario and the target's motion characteristics. "Absolute motion state" refers to whether the target is stationary or dynamic relative to the seabed; a stationary state indicates that the target's position remains essentially unchanged, while a dynamic state indicates that the target's position changes over time. "Qualitative processing" refers to the preliminary judgment and classification of the target's nature based on its absolute motion state, in order to adopt different processing strategies subsequently. "Navigation mark identification process" refers to the activation of specialized identification algorithms and processes for targets identified as navigation marks to confirm their type, location, and other detailed information. "Non-navigation mark dynamic obstacles" refers to targets identified as dynamic and not navigation marks; these targets typically require collision avoidance measures from the ship.

[0026] The core of the collision avoidance method for ship operations proposed in this application lies in achieving intelligent identification and collision avoidance decision-making for maritime targets through accurate judgment of the absolute motion state of the target and qualitative processing based on this.

[0027] First, it is necessary to acquire the target's absolute motion data relative to the seabed within a preset time period. This can be achieved in several ways. For example, a shipborne radar system can be used to continuously track the target, calculate the target's position and velocity relative to the ship using radar echo data, and then combine this with the ship's own GPS data and attitude sensor data to convert the target's motion data relative to the ship into absolute motion data relative to the seabed. Another approach is to use shipborne visual sensors, employing image processing and target tracking algorithms to identify and track the target, and then combine visual ranging and the ship's own motion data to estimate the target's absolute motion data. Furthermore, a sonar system can be used to acquire target motion data in the underwater environment. These sensors can operate independently or undergo data fusion to improve the accuracy and reliability of the motion data. For instance, in practical operation, a preset time period of 5 minutes can be set, during which the radar system acquires and records the target's position information every 10 seconds.

[0028] Secondly, based on the acquired absolute motion data, the absolute motion state of the target relative to the seabed is determined. This includes a static or dynamic state. The determination process can be based on the analysis of the target's position changing over time. For example, the average velocity and displacement of the target over a preset time period can be calculated. If the target's average velocity is below a preset static threshold and its total displacement is within a preset static range, the target can be determined to be in a static state. Conversely, if the target's average velocity or total displacement exceeds the corresponding threshold, the target is determined to be in a dynamic state. For example, if a target's average velocity relative to the seabed is less than 0.1 m / s and its total displacement is less than 3 m over 5 minutes, it can be determined to be in a static state.

[0029] Secondly, the target is qualitatively classified based on its absolute motion state. This step is one of the key innovations of this application. Specifically, if the target is stationary, the qualitatively classified target information initiates the navigation mark identification process. This means that the system initially considers the stationary target as a potential navigation mark and initiates a specialized algorithm and database for further identification and confirmation. For example, the system can call the navigation mark feature library to compare the visual features, radar features, etc., of the stationary target with the features of known navigation marks to determine whether it is a navigation mark and its type. If the target is dynamic, the qualitatively classified target information is a non-navigation mark dynamic obstacle. This means that the system initially considers the dynamic target as an obstacle that needs to be avoided, and no longer identifies it as a navigation mark. For example, a target moving at a speed of 0.5 m / s will be directly classified as a non-navigation mark dynamic obstacle, and the system will no longer attempt to match it with a navigation mark.

[0030] Finally, based on the qualitative target information, collision avoidance control commands are generated for the vessel. If the target is identified as a navigational aid and confirmed as a genuine navigational aid through the navigational aid identification process, the system will generate corresponding navigation commands based on the location and type of the navigational aid and the vessel's course plan, such as adjusting the course to remain within the channel. If the target is identified as a non-navigational aid dynamic obstacle, the system will calculate the optimal collision avoidance path based on the obstacle's position, speed, direction, and the vessel's own motion state, and generate corresponding collision avoidance control commands, such as adjusting the course, decelerating, or accelerating, to avoid a collision with the obstacle. For example, if the system identifies a non-navigational aid dynamic obstacle crossing the channel at a speed of 2 knots, it may generate a command requiring the vessel to turn 15 degrees to the right and decelerate by 1 knot to ensure safe passage.

[0031] The collision avoidance method for ship operations proposed in this application forms a complete intelligent collision avoidance decision chain through the synergistic effect of the above steps. First, high-precision absolute motion data of the target is acquired through multi-sensor fusion, laying the foundation for subsequent judgments. Second, based on the absolute motion data, the target's static or dynamic state is determined, avoiding the problem of misidentifying moving fishing gear as fixed navigation marks in traditional methods. Third, the target is qualitatively processed according to its motion state, initially distinguishing between potential navigation marks and dynamic obstacles, greatly improving the efficiency and accuracy of subsequent processing. Finally, collision avoidance control commands are generated based on the qualitatively determined target information, ensuring the safe navigation of the ship in complex maritime environments.

[0032] Compared with existing technologies, the core innovation of this application lies in the introduction of a qualitative processing mechanism based on the absolute motion state of the target. Traditional methods often attempt to identify all targets directly, leading to the misidentification of numerous non-standardized, dynamic fishing gear buoys as navigation marks in areas with intensive fishing activities, or the underreporting of genuine navigation marks due to visual noise. This influx of ambiguous and misleading visual information significantly reduces the system's confidence in navigable channels and may cause the automated control system to perform collision avoidance operations based on erroneous information, leading the vessel towards a truly dangerous area.

[0033] This application avoids misidentifying dynamic non-navigational objects as navigational aids by first determining whether the target is stationary or dynamic before identification. For example, a fishing buoy drifting with the current might be incorrectly identified as a fixed navigational aid in a traditional system, causing system confusion and potentially leading to incorrect collision avoidance maneuvers. However, the method in this application, because the buoy is dynamic, is directly classified as a non-navigational aid dynamic obstacle, thus avoiding the subsequent navigational aid identification process and directly entering the collision avoidance decision-making stage, significantly improving the accuracy and efficiency of the decision. Furthermore, this application only initiates the navigational aid identification process for stationary targets, allowing the navigational aid identification system to focus on objects that are truly likely navigational aids, reducing the false alarm rate and increasing the confidence in identifying genuine navigational aids. This phased, motion-state-based qualitative processing significantly enhances the vessel's environmental perception capabilities and the robustness of collision avoidance decisions in complex maritime environments, effectively reducing the risk of collisions, groundings, or other navigational accidents.

[0034] In some embodiments described above, a method for qualitatively classifying targets based on their absolute motion state relative to the seabed was proposed. However, in its implementation, relying solely on static or dynamic states for qualitative classification may not adequately address the complex and ever-changing marine environment. For example, some static targets may not be navigational aids, while some dynamic targets may possess special attributes, leading to uncertainties or risks of misjudgment in the qualitative results. To address this, this application further proposes a more refined qualitative classification method. This method calculates the confidence level of the target's motion state by comprehensively evaluating multi-dimensional information, thereby achieving more accurate target classification. Furthermore, an enhanced observation process is introduced to handle uncertainties.

[0035] In this regard, refer to Figure 2 S3 includes: S31, Evaluate the intensity of local environmental disturbance and the stability of local water surface texture around the target; S32, assess the correlation between the target and the electronic chart navigation mark; S33, calculate the confidence level of the target motion state based on the absolute motion data, the intensity of the local environmental disturbance, the stability of the local water surface texture, and the correlation degree of the electronic nautical chart beacon; S34, perform qualitative processing on the target based on the confidence level of the target motion state; The qualitative processing includes: when the confidence level of the target's motion state is higher than a first threshold, initiating the navigation mark identification process; when the confidence level of the target's motion state is lower than a second threshold, classifying the target as a non-navigation mark dynamic obstacle; and when the confidence level of the target's motion state is between the first threshold and the second threshold, initiating the enhanced observation process.

[0036] Specifically, assessing the intensity of local environmental disturbances around the target refers to quantitatively analyzing environmental factors (e.g., wave height, current speed, wind speed, etc.) within the local area where the target is located to determine the potential impact of these factors on the target's motion or stationary state. For example, real-time data can be obtained through shipborne sensors (such as wave sensors, current meters, and anemometers), or assessments can be conducted in conjunction with meteorological and sea state forecast data. The aim is to distinguish whether the target is inherently stationary or in motion, or whether its apparent stationaryness or motion is due to environmental factors.

[0037] Assessing the stability of local water surface textures involves analyzing images of the water surface surrounding the target to determine the uniformity, continuity, and trend of the texture. For example, visual sensors can be used to acquire water surface images, and image processing algorithms (such as texture analysis and optical flow methods) can be used to quantify the stability of the water surface texture. Stable water surface textures may indicate that the target is a fixed structure or is in calm water, while unstable textures may indicate that the target is moving or is disturbed by local water flow.

[0038] Assessing the correlation between the target and navigational aids on the electronic nautical chart involves comparing the target's location, shape, size, and other information with known navigational aid data on the electronic chart to determine the degree of match. For example, Geographic Information System (GIS) algorithms can be used to calculate the distance and direction between the target and the nearest navigational aid, and then compare their type characteristics. The purpose is to verify whether the target is a registered navigational aid, thereby avoiding misidentification of other stationary objects as navigational aids.

[0039] The target motion state confidence score can be understood as a comprehensive probability or score used to quantify the likelihood that the target is a navigational aid or a non-navigational aid dynamic obstacle. This confidence score is calculated by fusing multi-source information such as absolute motion data, local environmental disturbance intensity, local water surface texture stability, and navigational aid correlation on electronic charts. For example, methods such as Bayesian networks, fuzzy logic, or machine learning models can be used to weightedly fuse and infer various input data, thereby outputting a confidence score value between 0 and 1. The purpose is to provide a more refined and reliable judgment criterion. The first threshold and the second threshold are preset numerical boundaries used to classify the target motion state confidence score. When the confidence score is higher than the first threshold, it indicates that the target is highly likely to be a navigational aid; when the confidence score is lower than the second threshold, it indicates that the target is highly likely to be a non-navigational aid dynamic obstacle; when the confidence score is between the first and second thresholds, it indicates that the target state is uncertain and requires further confirmation.

[0040] The enhanced observation process refers to the system automatically initiating a series of additional observation and analysis measures when the target state is uncertain, in order to obtain more information and eliminate uncertainty. For example, this may include adjusting the sensor perspective, increasing the data acquisition frequency, initiating high-resolution imaging, deploying auxiliary detection equipment, or conducting manual verification. Its purpose is to avoid misjudgments and improve the accuracy of decision-making under uncertain circumstances.

[0041] This application's solution incorporates multi-dimensional information, including the intensity of local environmental disturbances, the stability of local water surface textures, and the correlation with electronic chart navigation marks, and fuses this information with absolute motion data to calculate the confidence level of the target's motion state. This multi-source information fusion mechanism allows for a more comprehensive and in-depth understanding of the target's environment and its true attributes, rather than solely relying on its apparent static or dynamic state. For example, even if a target appears stationary for a short period, if the intensity of local environmental disturbances is high, the water surface texture is unstable, and the correlation with electronic chart navigation marks is low, the confidence level for classifying it as a navigation mark will decrease, thus avoiding misclassification. Conversely, even if a target exhibits slight motion, if its correlation with electronic chart navigation marks is high and the surrounding environment is stable, the confidence level for classifying it as a navigation mark will increase. By setting a first threshold and a second threshold, this application can quickly and accurately identify dynamic obstacles, both high-confidence and non-navigational, with agility. Meanwhile, for ambiguous situations where the confidence level is in the middle range, an enhanced observation process is initiated to actively acquire more information, thereby effectively solving the problems of misjudgment and uncertainty that may occur in traditional methods in complex environments.

[0042] Through the above technical solutions, this application can significantly improve the accuracy and robustness of target qualitative processing during ship collision avoidance operations. By comprehensively considering multi-dimensional information, the system can more accurately distinguish between real navigation marks and other stationary or dynamic obstacles, effectively reducing the risk of false alarms and missed alarms. Especially when the target state is uncertain, the introduction of an enhanced observation process enables the system to proactively take measures to obtain more information, avoiding erroneous judgments when information is insufficient. This provides a more reliable and safer basis for ship collision avoidance decisions, further improving the safety and efficiency of ship navigation.

[0043] In some preferred embodiments, a specific example is given below. Suppose a ship is navigating when its detection system detects a target. First, the system acquires the target's absolute motion data relative to the seabed over a preset time period, initially determining that it is stationary. However, to avoid simply classifying it as a navigational aid, the system further assesses the intensity of local environmental disturbances around the target (e.g., moderate currents and slight waves detected by shipboard sensors), the stability of local water surface texture (e.g., slightly irregular water surface texture detected by visual sensors), and the correlation between the target and navigational aids on electronic charts (e.g., a slight deviation in its position from the nearest navigational aid on the electronic chart, and an incomplete shape match). Based on this multi-dimensional information, the system calculates the target's motion state confidence score, assuming this score is 0.55. If the preset first threshold is 0.7 and the second threshold is 0.3, then 0.55 falls between the first and second thresholds. In this uncertain situation, the system does not immediately classify the target as a navigational aid or a non-navigational aid dynamic obstacle, but instead initiates an enhanced observation process. For example, the system can automatically adjust the focal length and angle of the shipborne high-resolution camera to conduct continuous observation of the target for a longer period of time, and increase the radar scanning frequency to obtain more detailed images and motion trajectory data, thereby providing a more sufficient basis for subsequent accurate judgment.

[0044] In some embodiments of this application, an enhanced observation process is initiated when the confidence level of the target's motion state is between a first threshold and a second threshold. However, in its implementation, if the first and second thresholds are fixed, it may not be able to adequately adapt to the complex and ever-changing environmental conditions and target characteristics during ship operations. This could lead to insufficient accuracy in the qualitative processing of the target in some cases, or excessive initiation of the enhanced observation process, increasing the system burden.

[0045] In response, this application further proposes to initiate an enhanced observation process when the confidence level of the target motion state is between a first threshold and a second threshold, including: The ship's adjustment parameters are obtained, including its speed, current channel width, the importance level of the target on the electronic chart, and the local environmental disturbance index. Adjust the first threshold and the second threshold according to the adjustment parameters; The confidence level of the target motion state is evaluated based on the adjusted first threshold and second threshold, and the enhanced observation process is initiated based on the evaluated confidence level.

[0046] Specifically, acquiring the vessel's adjustment parameters aims to provide a multi-dimensional basis for threshold adjustment decisions. Among these, the ship's speed reflects its current operational status and required collision avoidance response time; the current channel width reflects the limitations of the ship's maneuverable space; the target's importance level on the electronic chart distinguishes the potential threat level of different targets—for example, large merchant ships and small fishing vessels may have different importance levels; and the local environmental disturbance index quantifies the complexity and uncertainty of the target's surrounding environment, such as the influence of wind, waves, and currents. These adjustment parameters can be acquired in real time by the ship's navigation system, environmental sensors, and electronic chart system.

[0047] Furthermore, adjusting the first and second thresholds according to the adjustment parameters means dynamically correcting the upper and lower limits used to determine the confidence level of the target's motion state based on the parameters obtained above. For example, when the ship's speed is high, the channel width is narrow, the target's importance level is high, or the local environmental disturbance index is large, the range between the first and second thresholds can be appropriately tightened, making the system more inclined to initiate enhanced observation procedures under conditions of high uncertainty, thereby improving the safety of collision avoidance decisions. Conversely, under favorable environmental conditions and low risk, the threshold range can be appropriately widened to reduce unnecessary enhanced observations. This adjustment can be achieved through preset rules, lookup tables, or machine learning-based models.

[0048] Therefore, based on the adjusted first and second thresholds, the confidence level of the target's motion state is reassessed, and the enhanced observation process is initiated based on the assessed confidence level. This means that after dynamically adjusting the thresholds, the system will again compare the calculated confidence level of the target's motion state with the new thresholds. If the confidence level is still between the adjusted first and second thresholds, it is confirmed that the enhanced observation process needs to be initiated; if the confidence level is higher than the adjusted first threshold or lower than the adjusted second threshold, the navigation mark identification process is initiated or the target is classified as a non-navigation mark dynamic obstacle, thereby achieving more refined qualitative processing.

[0049] This application's solution effectively addresses the problem in existing solutions where fixed thresholds can lead to inaccurate qualitative processing or over-activation of augmented observation processes by introducing ship adjustment parameters and dynamically adjusting the first and second thresholds of the target motion state confidence based on these parameters. Specifically, when the target motion state confidence is within an uncertain range, augmented observations are not simply activated. Instead, the boundaries of this uncertain range are redefined by considering the ship's own operational status (sailing speed, current channel width), the target's characteristics (importance level), and the complexity of the external environment (local environmental disturbance index). This adaptive threshold adjustment mechanism allows the system to more intelligently and flexibly determine whether augmented observations are needed based on the actual situation, thus avoiding unnecessary resource consumption in low-risk situations and enabling timely and more cautious strategies in high-risk situations.

[0050] Through the above technical solution, this application can significantly improve the adaptability and robustness of ship collision avoidance methods in complex and changing environments. By dynamically adjusting the threshold, the system can more accurately identify whether the target is a navigation mark or a dynamic obstacle, reducing the risk of misjudgment and missed judgment. This not only optimizes the timing and frequency of enhanced observation process activation and avoids waste of resources, but also provides more reliable decision-making basis at critical moments, thereby effectively improving the safety of ship navigation and reducing the probability of collision accidents.

[0051] In some preferred embodiments, a specific example is given below. Suppose a ship is sailing at a moderate speed in a channel of moderate width at night. The radar system detects a target with a calculated motion confidence score of 0.55, while the initially set first threshold is 0.6 and the second threshold is 0.4. Based on the initial thresholds, the target's motion confidence score of 0.55 falls between the first threshold of 0.6 and the second threshold of 0.4, therefore, an enhanced observation procedure is initiated.

[0052] However, according to the scheme of this application, before initiating the enhanced observation process, the system acquires the adjustment parameters of the vessel. For example, if the current sailing speed is 10 knots, the channel width is 500 meters, the importance level of the target on the electronic chart is medium, and the local environmental disturbance index is low (e.g., calm seas). Based on these parameters, the system dynamically adjusts the first threshold and the second threshold according to preset adjustment rules or models. For example, due to favorable environmental conditions and low risk, the system may adjust the first threshold to 0.58 and the second threshold to 0.42.

[0053] At the adjusted threshold, the target motion state confidence level of 0.55 is now lower than the adjusted first threshold of 0.58, but still higher than the adjusted second threshold of 0.42. Therefore, the system will reassess based on the adjusted threshold and confirm that the enhanced observation process still needs to be initiated.

[0054] To give another example, if a ship is traveling at high speed in a narrow channel and the local environmental disturbance index is high (e.g., high winds and waves), the adjustment parameters acquired by the system will prompt a more stringent adjustment of the thresholds. For example, the first threshold might be adjusted to 0.65, and the second threshold to 0.35. If the target motion state confidence level is still 0.55, then it will fall between the adjusted first threshold of 0.65 and the second threshold of 0.35, and the system will initiate an enhanced observation process. This dynamic adjustment ensures that the system can make more reasonable and safer decisions under different risk levels, avoiding over-observation at low risk and missing critical information at high risk due to threshold insensitivity.

[0055] In some embodiments described above, this application proposes calculating the confidence level of a target's motion state based on absolute motion data, local environmental disturbance intensity, local surface texture stability, and electronic chart beacon correlation. However, in practical applications, the reliability of these input data may vary due to differences in environmental conditions, sensor performance, or data acquisition methods. If the reliability of these data is not considered and fixed weights are directly assigned for confidence level calculation, the assessment of the target's motion state confidence level may be inaccurate or ineffective, thus affecting the accuracy of subsequent qualitative processing. Therefore, this application further proposes a method to optimize the calculation of the target's motion state confidence level by dynamically adjusting the weights of each input parameter to improve the accuracy of the confidence level assessment.

[0056] The steps described above for calculating the confidence level of the target's motion state based on absolute motion data, local environmental disturbance intensity, local water surface texture stability, and electronic chart and navigation mark correlation include: dynamically adjusting the corresponding weights based on the reliability of the absolute motion data, the reliability of the local environmental disturbance intensity, the reliability of the local water surface texture stability, and the reliability of the electronic chart and navigation mark correlation; and calculating the confidence level of the target's motion state based on the adjusted weights.

[0057] Specifically, "reliability of absolute motion data" refers to the trustworthiness of the acquired target absolute motion data in terms of accuracy, completeness, and timeliness. For example, when target motion data comes from multiple sensors and is mutually verified to be consistent, its reliability is high; conversely, if the data is conflicting or missing, the reliability is low. "Reliability of local environmental disturbance intensity" refers to the credibility of the assessment results of the local environmental disturbance intensity around the target, which may be affected by factors such as sensor measurement accuracy and environmental complexity. "Reliability of local water surface texture stability" refers to the credibility of the assessment results of the water surface texture stability in the target's area; for example, in severe sea conditions, water surface texture may be difficult to capture accurately, leading to reduced reliability. "Reliability of electronic chart beacon correlation" refers to the credibility of the degree of matching between the target and beacon information on the electronic chart, which may be affected by factors such as the electronic chart data update frequency and target identification accuracy.

[0058] "Dynamically adjusting the corresponding weights" refers to adjusting the contribution of each input parameter in calculating the confidence level of the target motion state in real time based on the reliability assessment results mentioned above. For example, when a parameter has high reliability, it can be given a larger weight; when its reliability is low, it can be given a smaller weight, or even reduced to near zero in extreme cases, to reduce its negative impact on the final confidence level calculation.

[0059] "Calculate the confidence level of the target motion state based on the adjusted weights" means that after obtaining the dynamically adjusted weights, these weights are applied to various input parameters, and the final confidence level of the target motion state is calculated by weighted averaging or other fusion algorithms.

[0060] This application's solution addresses the problem of inaccurate confidence assessment caused by inconsistent data reliability in traditional methods by introducing an evaluation of the reliability of each input data point and dynamically adjusting its weight in confidence calculation. Specifically, when the reliability of an input parameter (e.g., absolute motion data) is high, it is given greater influence in the confidence calculation of the target motion state, making the final confidence score more reflective of the state indicated by the high-reliability data. Conversely, when the reliability of an input parameter (e.g., local water surface texture stability) is low, its influence on the confidence calculation is weakened, thus avoiding interference from low-quality data in the overall assessment. This dynamic adjustment mechanism enables the confidence calculation to adaptively respond to changes in data quality under different environmental and sensor conditions, ensuring the accuracy and robustness of the confidence assessment.

[0061] Through the above technical solution, this application can significantly improve the accuracy and robustness of target motion state confidence calculation. By considering the reliability of each input data point and dynamically adjusting the weights, the system can more intelligently handle uncertainty and reduce misjudgments caused by poor quality of single or partial data. This enables ships to make more accurate and safer qualitative processing and collision avoidance control commands based on more reliable target motion state confidence when making collision avoidance decisions, effectively reducing the risk of false alarms and missed alarms, and improving the overall safety of ship operations.

[0062] In some preferred embodiments, it is assumed that the ship is navigating in dense fog. In this situation, the local water surface texture stability data acquired by visual sensors may be less reliable due to low visibility, while the absolute motion data acquired by the radar system may be relatively reliable. In this case, the system will assess the reliability of the local water surface texture stability as lower and correspondingly reduce its weight in the target motion state confidence calculation; simultaneously, the reliability of the absolute motion data is assessed as higher, and its weight is correspondingly increased. Through this dynamic adjustment, even under adverse weather conditions, the target motion state confidence can still be calculated primarily based on more reliable data, such as relying more on the absolute motion information provided by radar, thereby avoiding misjudgments caused by unreliable visual data and ensuring the accuracy of collision avoidance decisions.

[0063] In some of the embodiments described above in this application, although the weights are dynamically adjusted based on the reliability of various data to calculate the confidence level of the target motion state, a detailed and robust mechanism is not provided for obtaining the crucial "reliability of absolute motion data." If the reliability assessment of the absolute motion data is inaccurate or susceptible to interference, it may lead to distortion in subsequent weight adjustments, thereby affecting the accuracy of the confidence level of the target motion state and ultimately potentially causing deviations in collision avoidance decisions.

[0064] In response, this application further proposes a method for obtaining the reliability of the absolute motion data before dynamically adjusting the corresponding weights.

[0065] The reliability of acquiring the absolute motion data includes: Acquire the first estimate of the target motion data from the visual sensor; Acquire a second estimate of the target motion data from the radar system; Determine whether there is a conflict between the first estimate and the second estimate; When the conflict exists, the health status of the visual sensor and the radar system is obtained, and the corresponding weights are adjusted according to the health status. The extent to which current environmental parameters affect the visual sensor and the radar system is determined. Based on the degree of impact, the weights are adjusted again; The reliability of the absolute motion data is calculated based on the adjusted weights.

[0066] Specifically, the visual sensor can be understood as a device capable of acquiring visual or distance information about a target, such as a shipborne camera, lidar, or infrared sensor. Its first estimation of target motion data refers to extracting information such as the target's motion trend, speed, and direction from the raw sensor data using visual processing algorithms or point cloud processing algorithms. The radar system can be understood as a shipborne navigation radar or collision avoidance radar, etc. Its second estimation of target motion data refers to calculating the target's motion parameters from the radar echo using radar signal processing technology.

[0067] Determining whether there is a conflict between the first estimate and the second estimate involves comparing the estimated values ​​of target motion data from different types of sensors. For example, it involves comparing whether the differences between them in key parameters such as position, velocity, and heading exceed a preset threshold. When these differences are significant, a conflict is considered to exist, indicating that the data from at least one sensor may have bias or uncertainty.

[0068] When the aforementioned conflict exists, obtaining the health status of the visual sensor and the radar system refers to assessing the internal operating status of the sensors, such as sensor calibration, internal fault indication, noise level, and data integrity. Adjusting the corresponding weights based on the health status means that if a sensor is judged to be in poor health, the weight of its data in subsequent reliability calculations will be reduced to minimize its negative impact on the final result.

[0069] The degree of influence of the acquired current environmental parameters on the visual sensor and the radar system refers to considering the impact of external environmental factors (such as visibility, precipitation, sea clutter, wind, and waves) on sensor performance. For example, dense fog or heavy rainfall can severely affect the performance of the visual sensor, while sea clutter may interfere with the radar system's detection. The degree of influence can be assessed based on a preset environmental model or real-time environmental monitoring data. Adjusting the weights based on this degree of influence means, in addition to considering the sensor's own health status, further refining the weights of each sensor's data according to the actual impact of the environment on sensor performance, to ensure that the most reliable motion data is obtained under different environmental conditions.

[0070] Finally, the reliability of the absolute motion data is calculated based on the adjusted weights. This means fusing the first and second estimates after weight adjustment, for example, through weighted averaging, Kalman filtering, or other data fusion algorithms, to obtain a comprehensive and more confident absolute motion data reliability assessment result.

[0071] This application's solution effectively addresses the limitations and uncertainties that may arise from single-sensor data by introducing a multi-sensor data fusion and dynamic weight adjustment mechanism. First, by simultaneously acquiring estimation data from visual sensors and radar systems, redundant acquisition and cross-validation of target motion data are achieved. When data from these two different sensors conflict, the system can promptly detect potential data anomalies. Second, by further evaluating the health status of the sensors, data unreliability caused by sensor malfunctions or performance degradation can be identified, and their weights in data fusion can be adjusted accordingly, thereby reducing the negative impact of faulty sensors on the overall reliability assessment. Third, considering the varying impacts of the external environment on the performance of different sensors, this application further introduces an assessment of the influence of environmental parameters on sensors and adjusts the weights again. This allows for the priority acceptance of sensor data less affected by the environment under harsh environmental conditions, or the deweighting of more affected sensor data. It is precisely this multi-dimensional, dynamic weight adjustment strategy that enables the final calculated absolute motion data reliability to more accurately reflect the true motion state of the target, thus providing high-quality input for subsequent target motion state confidence calculations.

[0072] Through the above technical solutions, this application can significantly improve the accuracy and robustness of absolute motion data reliability assessment. Specifically, by fusing multi-source heterogeneous sensor data and performing collision detection, misjudgments caused by single sensor failure or environmental interference are effectively avoided. By dynamically assessing the sensor health status and the degree of environmental influence and adjusting the weights accordingly, the system can intelligently adapt to changes in sensor performance and the complex and ever-changing marine environment, ensuring high-confidence absolute motion data reliability under various operating conditions. This more accurate and reliable absolute motion data reliability will directly optimize the aforementioned dynamic weight adjustment process, thereby improving the calculation accuracy of the target motion state confidence, ultimately making the ship's collision avoidance control commands more accurate, timely, and safe, effectively reducing the risk of misjudgment and improving the safety of ship operations.

[0073] In some preferred embodiments, a specific example is given below. Suppose a ship is sailing at night and encounters localized dense fog. At this time, the ship's onboard visual sensor makes a first estimate of the motion data of a target ahead (e.g., a small fishing boat). Due to the fog, the estimated target speed is 5 knots and the heading is 090 degrees. Simultaneously, the ship's onboard radar system makes a second estimate of the same target, estimating its speed to be 7 knots and its heading to be 095 degrees.

[0074] First, the system determines that there is a conflict between the first estimate and the second estimate because there are significant differences in speed and heading.

[0075] Next, the system acquires the health status of the visual sensor and the radar system. Assuming the visual sensor is in good health, but its performance curve in low visibility shows a significant drop in accuracy; and the radar system is also in good health and has strong penetration capability in fog. Based on the health status, the weights are initially adjusted, for example, giving the radar system a higher initial weight.

[0076] The system then acquires the current environmental parameters, namely the localized dense fog, and assesses its impact on the visual sensor and the radar system. The impact of the dense fog on the visual sensor is assessed as high, while its impact on the radar system is assessed as low to medium.

[0077] Based on the degree of impact, the system readjusts the weights. Since dense fog has a greater impact on visual sensors, its weight is further reduced; while the radar system, although also affected to some extent, is relatively less affected, and its weight remains relatively high.

[0078] Finally, based on the adjusted weights, such as a visual sensor weight of 0.2 and a radar system weight of 0.8, the reliability of the absolute motion data is calculated. Through weighted averaging, the final absolute motion data is likely to be closer to the radar's estimate, such as a speed of 6.6 knots and a heading of 094 degrees, with a high reliability index. This reliability data, after multi-sensor fusion and dynamic weight adjustment, will be used for subsequent target motion state confidence calculations, thereby ensuring accurate collision avoidance decisions even in complex environments.

[0079] In some embodiments described above in this application, a method for obtaining reliable absolute motion data is proposed. This method assesses reliability by fusing estimation results from visual sensors and radar systems, and considering sensor health status and environmental influences. However, in actual shipboard operating environments, targets may temporarily move out of the direct observation range of sensors for various reasons (e.g., obstruction by other large ships, land structures, or adverse weather conditions), leading to data interruption or significant degradation in quality. In such cases, relying solely on real-time sensor data for reliability assessment may not accurately reflect the true reliability of target motion, and could even lead to misjudgment of the target's motion state.

[0080] In this regard, this application further proposes steps for obtaining the reliability of the absolute motion data, including: Acquire the first motion data of the target before occlusion, the first motion data including motion trend information, motion speed information, motion direction information, motion acceleration information and motion angular velocity information; Acquire second motion data of the target in the area where the shading occurs, the second motion data including local water flow direction information, local water flow velocity information, local wind direction information, and local wind speed information; Based on the first motion data and the second motion data, predict the possible range of motion of the target during the occlusion period; After the occlusion is removed, the third motion data of the target is acquired; Compare the third motion data with the possible motion range; The reliability of the absolute motion data is determined based on the comparison results.

[0081] Specifically, acquiring the target's initial motion data before obstruction refers to continuously collecting and analyzing its motion characteristics before the target enters or has already entered the sensor obstruction area. This initial motion data encompasses key information such as the target's motion trend, velocity, direction, acceleration, and angular velocity under normal observation conditions. This data provides crucial historical data for predicting the target's trajectory during obstruction. For example, continuous monitoring and data recording can be performed when the target is not obstructed using the ship's own navigation system, radar system, or visual sensors.

[0082] Acquiring the second motion data of the target in the area of ​​obstruction refers to collecting dynamic parameters of the local environment during the period when the target is obstructed. This second motion data mainly includes local water flow direction, local water flow velocity, local wind direction, and local wind speed. These environmental parameters have a significant impact on the motion of surface targets, especially when the target's own power is limited or it is floating. For example, this data can be acquired through shipborne weather stations, water flow sensors, or by combining environmental models with electronic nautical charts.

[0083] In practical applications, the possible range of motion of the target during occlusion is predicted based on the first motion data and the second motion data. This prediction process combines the target's inertia before occlusion (reflected by the first motion data) and the environmental influence within the occlusion area (reflected by the second motion data). Kalman filtering, particle filtering, neural networks, or other motion prediction models can be used to comprehensively consider the target's inertial motion and environmental disturbances, generating a probabilistic range of motion rather than a single predicted trajectory to address uncertainty.

[0084] After the obstruction is removed, the third motion data of the target is acquired. This third motion data refers to the target's actual motion data, including its position, velocity, and direction, acquired immediately after it re-enters the sensor's observable range. This can be achieved by reactivating the visual sensor, radar system, or other detection equipment.

[0085] Subsequently, the third motion data is compared with the possible motion range. This comparison aims to assess the consistency between the actual motion of the target and the predicted motion. For example, the deviation between the third motion data and the center point of the predicted motion range can be calculated, or it can be determined whether the third motion data falls within the predicted possible motion range.

[0086] Finally, the reliability of the absolute motion data is determined based on the comparison results. If the third motion data closely matches the predicted possible motion range, it indicates that the prediction of the target motion during the occlusion period is accurate, and thus the absolute motion data during this period can be considered to have high reliability. Conversely, if the deviation is large, it may be necessary to reduce the reliability assessment of the data during this period, or even trigger further anomaly detection or reassessment procedures.

[0087] The reliability of the absolute motion data obtained by the scheme in this application is due to its thorough consideration of the target's motion characteristics in complex environments, especially under occlusion. By acquiring detailed first motion data of the target before it is occluded, the system can establish a preliminary understanding of the target's motion inertia. Subsequently, during the period of target occlusion, by acquiring local environmental parameters of the occluded area (second motion data), the system can take into account the influence of environmental factors on the target's motion. It is precisely because of the combination of the target's own motion trend and the disturbance of the external environment that the system can make more accurate and robust predictions of the target's motion during occlusion, generating a reasonable possible motion range. When the target becomes visible again, the system can verify the accuracy of the prediction by comparing the actually observed third motion data with the predicted possible motion range. This verification mechanism effectively compensates for the inability of sensors to directly observe during occlusion, thereby enabling an objective and quantitative assessment of the reliability of the absolute motion data during occlusion.

[0088] Through the above technical solution, this application can significantly improve the accuracy and robustness of absolute motion data reliability assessment in scenarios where the target is occluded. Compared with methods that rely solely on real-time sensor data or simple extrapolation, this application constructs a more comprehensive reliability assessment model by fusing motion data before occlusion, environmental data of the occluded area, and verification data after occlusion. This not only avoids reliability assessment distortion caused by data interruption or quality degradation due to occlusion, but also effectively distinguishes between normal and abnormal target motion during occlusion, thereby providing more reliable and continuous target motion information for subsequent ship collision avoidance decisions, reducing the risk of misjudgment, and improving the safety of ship operations.

[0089] In some preferred embodiments, a specific example is given below. Suppose an autonomous vessel is performing a collision avoidance mission, and there is a target vessel ahead of it. When the target vessel is about to enter the rear of a large oil tanker, causing it to be blocked by the oil tanker, the solution of this application will be activated.

[0090] First, before the target vessel is obscured by the oil tanker, the ship's visual sensors and radar systems continuously monitor the target vessel, acquiring its initial motion data, including its speed, heading, acceleration, and turning rate. For example, it recorded the target vessel sailing stably at a speed of 10 knots and a heading of 270 degrees.

[0091] Secondly, when the target vessel enters the rear of the tanker and is blocked, the ship's local environmental sensors and weather station will acquire secondary motion data of the blocked area. For example, the water flow direction in the area is 280 degrees, the water flow speed is 0.5 knots, the wind direction is 260 degrees, and the wind speed is 10 meters per second.

[0092] Next, the system uses this first and second motion data, combined with a motion prediction model (e.g., a physical model that considers the effects of water flow and wind), to predict the possible range of motion of the target vessel during the period of obstruction. For example, it predicts that it may have a heading range of 270 to 285 degrees and a speed between 9.5 and 10.5 knots.

[0093] Subsequently, once the target vessel sailed out from the other side of the tanker and the obstruction was lifted, the ship's sensors immediately acquired the target vessel's third motion data. For example, the target vessel was actually observed to continue sailing at a speed of 10.2 knots and a heading of 275 degrees.

[0094] Finally, the system compares the actual observed third motion data (10.2 knots, 275 degrees) with the predicted possible motion range (heading 270-285 degrees, speed 9.5-10.5 knots). Since the actual data falls within the predicted range and the deviation is small, the system determines that the absolute motion data of the target vessel during the obstruction period has high reliability. This method ensures that the reliability of the target motion can be continuously and accurately assessed even when the target is obstructed, thus providing a solid data foundation for the ship's collision avoidance decisions.

[0095] In some embodiments described above, this application proposes acquiring the first motion data of a target before occlusion. However, in practical applications, a target may undergo nonlinear motion before entering the occlusion area, such as abrupt turning or acceleration / deceleration. This makes it difficult to accurately capture its instantaneous motion state and future trend using only conventional motion data monitoring. This inaccuracy may lead to deviations in the prediction of the target's possible motion range during occlusion, thereby affecting the timeliness and effectiveness of collision avoidance decisions. To address this, this application further proposes a method for acquiring the first motion data of a target before occlusion. This method continuously monitors the target's rate of change of position, rate of change of velocity, and rate of change of direction, and employs short-time window analysis and tracking of instantaneous angular acceleration at turning points for nonlinear motion to more accurately acquire the target's motion data before occlusion.

[0096] The aforementioned acquisition of the target's first motion data before occlusion specifically includes: The system continuously monitors the target's rate of change of position, rate of change of velocity, and rate of change of direction. Continuous monitoring refers to the system acquiring the target's position, velocity, and direction information at a preset frequency or in real-time, and calculating the rate of change of these parameters over time. The rate of change of position reflects the target's movement trend, the rate of change of velocity indicates whether the target is accelerating or decelerating, and the rate of change of direction indicates whether the target is turning. By continuously monitoring these rates of change, the target's dynamic behavior can be comprehensively understood.

[0097] When the rate of change of the target's velocity or the rate of change of its direction exceeds a preset threshold, the target is determined to be undergoing nonlinear motion. Specifically, the preset threshold is determined comprehensively based on factors such as the typical motion characteristics of the ship, sensor accuracy, and collision avoidance safety requirements. For example, when the rate of change of the target's velocity (such as acceleration) exceeds a certain value, or the rate of change of its direction (such as angular velocity) exceeds a certain value, the target can be considered to be undergoing nonlinear motion, rather than simple linear uniform motion.

[0098] When the target is undergoing nonlinear motion, motion data within a short time window is used for analysis. This is because the target's motion characteristics may change significantly within a short period during nonlinear motion. Using a short time window analysis allows for more timely and precise capture of these instantaneous changes, avoiding information distortion caused by long-term window averaging, thereby improving the accuracy of perception of the current motion state.

[0099] The system tracks the instantaneous angular acceleration of a target at a turning point and predicts the trajectory around that point. Turning points are typically critical moments where the target's direction or velocity changes significantly. Tracking instantaneous angular acceleration allows for precise quantification of the target's dynamic characteristics during turns. By combining the motion trend before and after the turning point with the instantaneous angular acceleration, a more accurate prediction of the target's future trajectory around the turning point can be made, providing a reliable basis for subsequent collision avoidance decisions.

[0100] This application's solution, by introducing continuous monitoring of the target's rate of motion change, enables real-time perception of the target's motion trend. When the rate of change of velocity or direction exceeds a preset threshold, the system can promptly identify the nonlinear motion the target is undergoing, thus avoiding the limitations of relying solely on linear models for prediction in complex motion scenarios. Furthermore, for nonlinear motion, short-time-window motion data is used for analysis, enabling the system to capture instantaneous and subtle motion changes, improving the accuracy of perception of the target's current motion state. Especially when the target is performing turning or other turning motions, by tracking its instantaneous angular acceleration, the dynamic characteristics of the turn can be more accurately characterized, and predictions can be made by combining historical data around the turning point. This significantly improves the accuracy and reliability of the initial motion data acquired before occlusion occurs, providing a solid foundation for subsequent prediction of the target's possible motion range during occlusion.

[0101] Through the above technical solution, this application effectively solves the problem of traditional methods struggling to accurately acquire motion data when a target is undergoing nonlinear motion. Specifically, by continuously monitoring the rate of change of motion in multiple dimensions and combining it with a threshold judgment mechanism, timely identification of the target's nonlinear motion is achieved. Based on this, short-time window analysis and instantaneous angular acceleration tracking are employed, significantly improving the accuracy and real-time performance of acquiring initial motion data in complex motion scenarios. This allows for more accurate prediction of the target's motion range during occlusion, significantly reducing the risk of collision avoidance decisions due to inaccurate motion data, thereby enhancing the overall safety and reliability of ship collision avoidance systems.

[0102] In some preferred embodiments, suppose a target ship is approaching an area obscured by a large oil tanker. Before entering the obscured area, the target ship suddenly makes a sharp right turn. According to the solution of this application, the shipborne system continuously monitors the target ship's rate of change of position, rate of change of velocity, and rate of change of direction. When the system detects that the target ship's rate of change of direction (e.g., angular velocity) exceeds a preset threshold (e.g., 5 degrees per second), the system immediately determines that the target ship is undergoing nonlinear motion. At this time, the system no longer uses average data over a longer time window, but switches to analyzing motion data over a shorter time window (e.g., 1-2 seconds) to capture its instantaneous turning characteristics. Simultaneously, the system tracks the target ship's instantaneous angular acceleration at the turning point and, combined with the motion trend before and after the turning point, predicts its possible course and speed after the turn is completed. In this way, even if the target ship undergoes complex nonlinear motion before the obstruction, the system can obtain highly accurate initial motion data, thereby providing reliable input for subsequent prediction of its motion range during the obstruction period, ensuring timely and accurate collision avoidance comparison and decision-making after the obstruction is lifted.

[0103] In some embodiments described above in this application, a method is proposed for tracking the instantaneous angular acceleration of a target at a turning point and predicting the area around the turning point. However, in its implementation, the accuracy of the instantaneous acceleration is easily affected by various factors, such as dynamically changing local environmental conditions, performance fluctuations of sensing devices such as visual sensors and radar systems, and the complexity of the target's own motion patterns. If the calculation of the instantaneous acceleration is not accurate or robust enough, especially when the target is making nonlinear motion (such as a sharp turn), it may lead to deviations in the prediction of the target's future trajectory, thereby affecting the timeliness and effectiveness of ship collision avoidance decisions and increasing the potential collision risk.

[0104] In response, this application further proposes a method to optimize the instantaneous acceleration of the tracking target at the turning point, which improves the accuracy and robustness of instantaneous acceleration calculation by comprehensively considering multiple factors.

[0105] The instantaneous acceleration of the tracked target at the inflection point includes: Get the current environment parameters; Based on the current environmental parameters, assess the extent to which the current environmental parameters affect the performance of the visual sensor and the radar system; Obtain the historical acceleration trend of the target before and after the inflection point; Compare the current instantaneous acceleration with the historical acceleration trend for consistency; Adjust the weight of instantaneous acceleration assessment based on the consistency comparison results; Based on the adjusted weights, the consistency comparison results, and the degree of influence, the instantaneous acceleration of the target at the inflection point is calculated.

[0106] Specifically, current environmental parameters refer to the external conditions affecting the target's motion or sensor performance near the turning point of a target's nonlinear motion (such as turning). These parameters may include, but are not limited to, visibility, precipitation intensity, sea clutter intensity, water current direction, water current velocity, wind direction, and wind speed. Obtaining these parameters helps to comprehensively understand the actual environment in which the target is located.

[0107] The degree of impact refers to the negative effects of current environmental parameters on the measurement accuracy, detection range, and signal quality of visual sensors and radar systems used to acquire target motion data. For example, high precipitation intensity or dense fog reduces the visibility of visual sensors, while strong sea clutter interferes with the echo signal of radar systems. Assessing this degree of impact is to quantify the reliability of sensor data under the current environment, in order to facilitate subsequent data fusion and weight adjustment.

[0108] In practical applications, historical acceleration trends refer to the regularity of a target's acceleration changes over a period of time, particularly when it experiences similar turning points. This can include the target's average acceleration, standard deviation of acceleration, and patterns of acceleration variation over a past period. By analyzing historical data, a predicted acceleration model or range can be established for comparison with the current instantaneous acceleration.

[0109] Specifically, consistency comparison refers to comparing the real-time instantaneous acceleration with an acceleration trend or model built based on historical data to determine whether the current value is within a reasonable range or whether there are any abnormal deviations. For example, if the current instantaneous acceleration is significantly inconsistent with the historical trend, it may indicate errors in the sensor data or abnormal movement of the target. Weight adjustment refers to dynamically allocating the importance of different data sources or evaluation methods in calculating the final instantaneous acceleration based on the results of the consistency comparison. If the current instantaneous acceleration is highly consistent with the historical trend, it can be assigned a higher weight; conversely, if there is a large deviation, its weight may need to be reduced, or it may need to be corrected by combining other information.

[0110] The proposed solution comprehensively considers the adjusted weights, consistency comparison results, and the influence of environmental parameters on sensor performance, and employs a data fusion algorithm (such as Kalman filtering, particle filtering, or other weighted averaging algorithms) to calculate a more accurate and reliable instantaneous acceleration of the target at the inflection point. The fusion process aims to maximize the utilization of effective information while suppressing noise and uncertainty.

[0111] Through the above technical solution, this application can significantly improve the accuracy and reliability of instantaneous acceleration calculation at turning points. Specifically, by considering the impact of current environmental parameters on sensor performance, measurement errors caused by harsh environments can be effectively avoided, ensuring the quality of input data. Simultaneously, introducing historical acceleration trends for consistency comparison can promptly detect and correct abnormal instantaneous acceleration data, enhancing the robustness of data processing. This multi-dimensional information fusion and dynamic weight adjustment mechanism makes the calculated instantaneous acceleration closer to the actual motion state of the target, especially when the target is undergoing nonlinear motion, providing a more accurate basis for motion prediction. Therefore, the overall reliability of absolute motion data is further improved, providing more solid and accurate data support for subsequent ship collision avoidance decisions, thereby effectively reducing collision risks and improving the safety of ship operations.

[0112] In some preferred embodiments, a specific example is given below. Suppose a ship is navigating in complex waters, and its visual sensors and radar systems are continuously monitoring a target that is making a turn.

[0113] First, the system acquires current environmental parameters. For example, the shipborne local environment sensor detects moderate rainfall and a certain intensity of sea clutter within the local environmental sensing area.

[0114] Next, based on these current environmental parameters, the system assesses the impact of rainfall and sea clutter on the performance of the visual sensor and radar system. For example, the assessment results show that rainfall reduced the visibility of the visual sensor by 20%, and sea clutter reduced the signal-to-noise ratio of the radar system by 15%.

[0115] Simultaneously, the system will acquire the historical acceleration trend of the target during similar turning maneuvers. For example, historical data shows that this type of target typically experiences an acceleration of 0.5 m / s² during a turn. 2 up to 1.5m / s 2 The instantaneous acceleration range is relatively small, and the acceleration changes are relatively stable.

[0116] At this moment, the sensor measures the instantaneous acceleration as 2.0 m / s². 2 The system compares this current instantaneous acceleration with historical acceleration trends. It detects 2.0 m / s². 2 If the value exceeds the upper limit of historical trends, there may be measurement errors or abnormal operation of the target.

[0117] Based on the consistency comparison results, the system adjusts the weights of instantaneous acceleration assessment. Because the current value deviates from historical trends, and environmental parameters negatively impact sensor performance, the system reduces the weight of real-time sensor data in calculating instantaneous acceleration and increases the weights of historical trend data and environmental impact assessment.

[0118] Finally, based on the adjusted weights, consistency comparison results, and the degree of environmental impact, the system integrates this information to calculate the instantaneous acceleration of the target at the inflection point. For example, after fusion calculation, the final determined instantaneous acceleration is 1.6 m / s². 2 This is higher than the original 2.0 m / s 2 It is closer to reality and takes into account environmental and historical experience adjustments. This more accurate instantaneous acceleration will be used for subsequent target motion prediction and collision avoidance control command generation.

[0119] In some embodiments described above in this application, in order to more accurately track the instantaneous acceleration of the target at the turning point, it is necessary to obtain the current environmental parameters. Specifically, obtaining the current environmental parameters includes the following steps: Obtain the ship's own position information; Obtain the position information of the target relative to the ship; Based on the ship's own position information and the target's position information relative to the ship, calculate the absolute position of the target relative to the seabed; A local environmental perception area is defined with the absolute position of the target relative to the seabed as the center; The shipborne local environment sensor is activated to scan the local environment sensing area and obtain the visibility, precipitation intensity, sea clutter intensity, water flow direction, water flow velocity, wind direction, and wind speed within the local environment sensing area to generate local environmental parameters of the target area.

[0120] Specifically, the ship's own position information can be obtained by the ship's Global Positioning System (GPS), Inertial Navigation System (INS), or other high-precision positioning equipment. This position information typically includes data such as the ship's longitude, latitude, altitude, and heading. The target's position information relative to the ship can be obtained using shipborne radar systems, visual sensors, LiDAR, and other equipment. These devices can monitor and output relative position data such as the target's distance and azimuth relative to the ship in real time.

[0121] Furthermore, after obtaining the ship's own position information and the target's position information relative to the ship, a coordinate transformation algorithm is used to convert the relative position information into absolute position information, thereby calculating the target's absolute position relative to the seabed. For example, the ship's own position can be used as a reference point, combined with the target's relative position, to obtain the target's precise position in the Earth coordinate system through vector superposition or geometric calculation.

[0122] Based on this, a local environmental sensing area can be defined with the absolute position of the target relative to the seabed as the center. The extent of this area can be dynamically adjusted according to actual application needs, target motion characteristics, and sensor detection capabilities. For example, it can be set as a circular or rectangular area with a radius of several hundred meters to several kilometers centered on the target. The purpose of defining the local environmental sensing area is to focus the acquisition of environmental parameters on the area that has the most direct impact on the target's motion.

[0123] Subsequently, the shipborne local environment sensors are activated to scan the local environment sensing area. These sensors may include, but are not limited to, weather radar, sonar, Doppler current meters, anemometers, and visibility meters. These sensors work together to collect real-time data on various environmental parameters within the local environment sensing area, such as visibility, precipitation intensity, sea clutter intensity, current direction, current velocity, wind direction, and wind speed. For example, a visibility meter measures atmospheric transparency, precipitation intensity can be obtained from weather radar, sea clutter intensity can be analyzed by a radar system, current direction and velocity can be measured by a Doppler current meter, and wind direction and speed are provided by anemometers. Through the fusion processing of these multi-source environmental data, the local environmental parameters of the target area are ultimately generated.

[0124] The proposed solution acquires detailed local environmental parameters of the target area, providing accurate and real-time input for subsequent assessment of the impact of current environmental parameters on the performance of visual sensors and radar systems. Specifically, by acquiring the ship's own position information and the target's position information relative to the ship, the absolute position of the target relative to the seabed can be accurately calculated, ensuring the accuracy of the local environmental perception area. By defining the perception area centered on the target's absolute position and scanning it using multiple shipborne local environmental sensors, key environmental factors such as visibility, precipitation intensity, sea clutter intensity, current direction, current speed, wind direction, and wind speed around the target can be comprehensively and meticulously captured. These detailed environmental parameters are crucial for accurately assessing sensor performance under specific conditions; for example, high precipitation intensity and sea clutter can significantly reduce radar detection capabilities, while low visibility can affect the performance of visual sensors.

[0125] The aforementioned technical solutions provide more comprehensive and accurate local environmental parameters, significantly improving the accuracy of sensor performance assessment. This precise acquisition of environmental parameters helps to more accurately understand and compensate for the impact of environmental factors on sensor data reliability, thereby improving the accuracy and robustness of target instantaneous acceleration tracking. Especially in complex and variable marine environments, it effectively avoids misjudgments of sensor performance caused by inaccurate environmental parameter acquisition, providing more reliable data support for ship collision avoidance decisions.

[0126] This application proposes a collision avoidance system for ship operations. This system achieves intelligent identification and collision avoidance decision-making for maritime targets through structured functional modules, effectively solving the accuracy problem of target identification and collision avoidance decision-making in complex maritime environments using traditional methods.

[0127] Reference Figure 3 This invention provides a schematic diagram of a ship collision avoidance system, including: The detection end is used to acquire the absolute motion data of the target relative to the seabed within a preset time period; The processing unit is used to determine the absolute motion state of the target relative to the seabed based on the absolute motion data, wherein the absolute motion state includes a static state or a dynamic state; and to perform qualitative processing on the target based on the absolute motion state. The control unit is used to generate collision avoidance control commands for the ship based on the qualitative target information. The qualitative processing includes: if the target is in the stationary state, the qualitative target information is used to initiate the navigation mark identification process; if the target is in the dynamic state, the qualitative target information is used to identify a non-navigation mark dynamic obstacle.

[0128] It should be noted that the ship operation collision avoidance system provided in this embodiment of the invention is used to execute all the process steps of the ship operation collision avoidance method in the above embodiment. The working principle and beneficial effects of the two are one-to-one, so they will not be described again.

[0129] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A method of collision avoidance for a marine vessel operation, characterized by, The method comprises: acquiring absolute motion data of a target relative to a seabed within a preset time; judging an absolute motion state of the target relative to the seabed according to the absolute motion data, the absolute motion state comprising a static state or a dynamic state; qualitatively processing the target based on the absolute motion state; generating a collision avoidance control instruction of a ship according to the qualitatively processed target information; wherein the qualitative processing comprises: if the target is in the static state, the qualitatively processed target information is starting a navigation mark identification process; if the target is in the dynamic state, the qualitatively processed target information is a non-navigation mark dynamic obstacle; and the control instruction at least comprises an instruction for controlling a current operation mode, an operation region boundary, and a key operation device.

2. A method of collision avoidance for a marine vessel according to claim 1, wherein, The qualitative processing of the target based on the absolute motion state comprises: evaluating local environmental disturbance intensity and local water surface texture stability around the target; evaluating an association degree of the target and an electronic chart navigation mark; calculating a target motion state confidence according to the absolute motion data, the local environmental disturbance intensity, the local water surface texture stability, and the electronic chart navigation mark association degree; qualitatively processing the target according to the target motion state confidence; wherein the qualitative processing comprises: when the target motion state confidence is higher than a first threshold value, starting the navigation mark identification process; when the target motion state confidence is lower than a second threshold value, qualitatively processing the target as the non-navigation mark dynamic obstacle; and when the target motion state confidence is between the first threshold value and the second threshold value, starting an enhanced observation process.

3. A method of collision avoidance for a marine vessel according to claim 2, wherein, When the target motion state confidence is between the first threshold value and the second threshold value, starting the enhanced observation process comprises: acquiring adjustment parameters of the ship, the adjustment parameters comprising a navigation speed, a current channel width, an importance level of the target on an electronic chart, and a local environmental disturbance index; adjusting the first threshold value and the second threshold value according to the adjustment parameters; evaluating the target motion state confidence according to the adjusted first threshold value and the second threshold value, and starting the enhanced observation process according to the evaluated confidence.

4. A ship maneuvering collision avoidance method according to claim 2, characterized in that, The calculation of the target motion state confidence according to the absolute motion data, the local environmental disturbance intensity, the local water surface texture stability, and the electronic chart navigation mark association degree comprises: dynamically adjusting corresponding weights according to reliability of the absolute motion data, reliability of the local environmental disturbance intensity, reliability of the local water surface texture stability, and reliability of the electronic chart navigation mark association degree; calculating the target motion state confidence according to the adjusted weights.

5. A ship maneuvering collision avoidance method according to claim 4, characterized in that, Before the dynamic adjustment of the corresponding weights, further comprising: acquiring the reliability of the absolute motion data; The acquisition of the reliability of the absolute motion data comprises: acquiring a first estimation of target motion data by a visual sensor; acquiring a second estimation of the target motion data by a radar system; judging whether the first estimation and the second estimation are in conflict; and When the conflict exists, a health status of the visual sensor and the radar system is acquired, and a corresponding weight is adjusted according to the health status; An influence degree of a current environmental parameter on the visual sensor and the radar system is acquired; The weight is adjusted again according to the influence degree; The reliability of the absolute motion data is calculated according to the adjusted weight.

6. A ship maneuvering collision avoidance method according to claim 5, characterized in that, The acquiring of the reliability of the absolute motion data comprises: First motion data of the target before the occlusion is acquired, the first motion data comprising motion trend information, motion speed information, motion direction information, motion acceleration information and motion angular velocity information; Second motion data of the target in an occlusion occurrence area is acquired, the second motion data comprising local water flow direction information, local water flow speed information, local wind direction information and local wind speed information; A possible motion range of the target during the occlusion is predicted according to the first motion data and the second motion data; After the occlusion is removed, third motion data of the target is acquired; The third motion data is compared with the possible motion range; The reliability of the absolute motion data is acquired according to a comparison result.

7. A ship maneuvering collision avoidance method according to claim 6, characterized in that, The acquiring of the first motion data of the target before the occlusion comprises: A position change rate, a speed change rate and a direction change rate of the target are continuously monitored; When the speed change rate or the direction change rate of the target exceeds a preset threshold, it is judged that the target is performing a nonlinear motion; When the target is performing the nonlinear motion, motion data of a short time window is used for analysis; Instantaneous angular acceleration of the target at a turning point is tracked and a surrounding of the turning point is predicted.

8. A ship maneuvering collision avoidance method according to claim 7, characterized in that, The tracking of the instantaneous acceleration of the target at the turning point comprises: A current environmental parameter is acquired; An influence degree of the current environmental parameter on performance of the visual sensor and the radar system is evaluated according to the current environmental parameter; A historical acceleration trend of the target before and after the turning point is acquired; A consistency comparison is made between a current instantaneous acceleration and the historical acceleration trend; A weight of instantaneous acceleration evaluation is adjusted according to a consistency comparison result; The instantaneous acceleration of the target at the turning point is calculated according to the adjusted weight, the consistency comparison result and the influence degree.

9. A ship maneuvering collision avoidance method according to claim 8, characterized in that, The acquiring of the current environmental parameter comprises: Ship self-position information is acquired; Position information of the target relative to the ship is acquired; An absolute position of the target relative to a seabed is calculated according to the ship self-position information and the position information of the target relative to the ship; A local environmental perception area is demarcated with the absolute position of the target relative to the seabed as a center; A shipborne local environmental sensor is started to scan the local environmental perception area, and visibility, precipitation intensity, sea clutter intensity, water flow direction, water flow speed, wind direction and wind speed in the local environmental perception area are acquired to generate local environmental parameters of an area where the target is located.

10. A ship maneuvering collision avoidance system characterized by, The system comprises: A detection end is configured to acquire absolute motion data of a target relative to a seabed within a preset time; The processing end is configured to determine an absolute motion state of the target relative to the seabed according to the absolute motion data, and the absolute motion state includes a static state or a dynamic state; and perform qualitative processing on the target based on the absolute motion state. The control end is configured to generate a collision avoidance control instruction of the ship according to the qualitative target information. The qualitative processing includes: if the target is in the static state, the qualitative target information is to start a navigation mark identification process; and if the target is in the dynamic state, the qualitative target information is a non-navigation mark dynamic obstacle.

Images