Method and system for detecting underwater foreign matters by ship based on visual detection

Abstract

The invention discloses a method and a system for detecting underwater foreign matters by a ship based on visual detection, and relates to the technical field of underwater detection methods of ships, and the method comprises the steps: determining each contact risk node based on the form of an underwater part of a ship; the navigation attitude of the ship is optimized according to the node position of each contact risk node and the distribution position of each underwater foreign matter, and the type of each underwater foreign matter is determined according to the form of each underwater foreign matter and the corresponding color parameter, so that the identification accuracy of the type of each underwater foreign matter is improved. Therefore, a ship navigation event is determined according to the underwater foreign matter event, the flow parameters of the water surface and the surrounding environment parameters of the ship; the multiple navigation risk factors of the ship are determined according to the ship navigation event and the previous navigation record, the underwater foreign matter processing mode of the ship is determined according to the multiple navigation risk factors, the underwater foreign matter event and the ship navigation mode, and the accuracy of the underwater foreign matter processing mode of the ship is improved.

Description

Technical Field

[0001] This invention relates to the technical field of underwater inspection methods for ships, and more particularly to a visual inspection-based method and system for detecting underwater foreign objects on ships. Background Technology

[0002] With the development of technology, ships refer to all kinds of transportation vehicles capable of navigating or operating on water, including but not limited to cargo ships, passenger ships, oil tankers, fishing boats, warships, engineering vessels, and research vessels. During navigation, ships may encounter underwater obstacles such as shipwrecks, reefs, fishing nets, and drifting debris. Failure to detect and avoid these obstacles in time can lead to damage to the hull, grounding, or even sinking. Current technologies analyze the morphology of underwater foreign objects by emitting sound waves and receiving echoes, but do not consider the color parameters of the underwater foreign objects. This affects the accuracy of identifying the types of underwater foreign objects, resulting in lower accuracy in underwater foreign object incident detection and impacting the precision of ship-based handling methods for underwater foreign objects. Summary of the Invention

[0003] The purpose of this invention is to overcome the shortcomings of the prior art. This invention provides a visual detection method and system for detecting underwater foreign objects by ships.

[0004] This invention provides a visual detection method for detecting underwater foreign objects (FOOs) on a ship, comprising: the ship navigating relative to the water surface; determining the ship's underwater control space based on the ship's navigation direction, draft, and the shape of the ship's underwater portion; determining image data based on visual detection of the underwater control space; determining the distribution locations of multiple FEOs based on acoustic data; determining the shape of each FEO based on the distribution locations of the multiple FEOs and the corresponding image data; determining each contact risk node based on the shape of the ship's underwater portion; optimizing the ship's navigation attitude based on the node locations of each contact risk node and the distribution locations of each FEO; determining the type of each FEO based on its shape and corresponding color parameters; determining an FEO event based on the actual number, distribution location, and type of each FEO; determining a ship navigation event based on the FEO event, water surface flow parameters, and ship's surrounding environmental parameters; determining multiple navigation risk factors for the ship based on the ship navigation events and previous navigation records; and determining the ship's handling method for the FEO based on the multiple navigation risk factors, the FEO event, and the ship's navigation mode.

[0005] This invention provides a vision-based detection system for underwater foreign objects (FMOs) on ships. This system is applied to the aforementioned vision-based FMO detection method. The vision-based FMO detection system includes: The underwater control space module is used for ship navigation relative to the water surface. It determines the ship's underwater control space based on the ship's navigation direction, draft, and the shape of the ship's underwater parts. The multimodal data module is used to determine image data based on visual inspection of the underwater control space, determine the distribution location of multiple underwater foreign objects based on acoustic data, and determine the morphology of each underwater foreign object based on the distribution location of multiple underwater foreign objects and the corresponding image data. The classification module is used to determine each contact risk node based on the shape of the underwater part of the ship, optimize the ship's navigation attitude based on the node position of each contact risk node and the distribution position of each underwater foreign object, and determine the type of each underwater foreign object based on the shape and corresponding color parameters. The ship navigation event module is used to determine underwater foreign object events based on the actual number, distribution location and type of each underwater foreign object, and to determine ship navigation events based on underwater foreign object events, water surface flow parameters and ship surrounding environment parameters. The handling method module is used to determine multiple navigation risk factors of a ship based on ship navigation events and past navigation records, and to determine the ship's handling method for underwater foreign objects based on multiple navigation risk factors, underwater foreign object events, and the ship's navigation mode.

[0006] Compared with the prior art, the beneficial effects of the present invention are: In this embodiment of the invention, the method is used to determine each contact risk node based on the morphology of the underwater part of the ship, optimize the ship's navigation attitude based on the node position of each contact risk node and the distribution position of each underwater foreign object, and determine the type of each underwater foreign object based on the morphology and corresponding color parameters of each underwater foreign object. The morphology of the underwater foreign object is introduced, and the overall consideration of the morphology and corresponding color parameters of each underwater foreign object is taken into account, thereby improving the accuracy of identifying the type of each underwater foreign object.

[0007] Therefore, underwater foreign object events are determined based on the actual quantity, distribution location, and type of each underwater foreign object. Ship navigation events are determined based on underwater foreign object events, surface flow parameters, and ship's surrounding environmental parameters. Multiple navigation risk factors for the ship are determined based on ship navigation events and past navigation records. The ship's handling method for underwater foreign objects is determined based on multiple navigation risk factors, underwater foreign object events, and the ship's navigation mode. By introducing ship navigation events, a holistic consideration of multiple navigation risk factors, underwater foreign object events, and the ship's navigation mode is achieved, improving the accuracy of the ship's handling method for underwater foreign objects. The ship has realized the detection and handling of underwater foreign objects. Attached Figure Description

[0008] Figure 1This is a flowchart illustrating a vision-based detection method for underwater foreign objects by a ship, as described in an embodiment of the present invention. Figure 2 This is a flowchart illustrating step S11 of the visual detection-based method for detecting underwater foreign objects by a ship in an embodiment of the present invention. Figure 3 This is a flowchart illustrating step S12 of the visual detection-based method for detecting underwater foreign objects by a ship in an embodiment of the present invention. Figure 4 This is a flowchart illustrating step S13 of the visual detection-based method for detecting underwater foreign objects by a ship in an embodiment of the present invention. Figure 5 This is a flowchart illustrating step S14 of the visual detection-based method for detecting underwater foreign objects by a ship in an embodiment of the present invention. Figure 6 This is a flowchart illustrating step S15 of the visual detection-based method for detecting underwater foreign objects by a ship in an embodiment of the present invention. Figure 7 This is a schematic diagram of the structural composition of a vision-based detection system for underwater foreign objects on a ship, according to an embodiment of the present invention. Detailed Implementation

[0009] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0010] Please see Figures 1 to 7 A vision-based detection method for detecting underwater foreign objects on ships, applied to underwater ship scenarios; the vision-based detection method for detecting underwater foreign objects on ships includes: Step S11: The ship navigates relative to the water surface. The underwater control space of the ship is determined based on the ship's navigation direction, draft, and the shape of the ship's underwater parts. Step S12: Determine image data based on visual detection of the underwater control space, determine the distribution location of multiple underwater foreign objects based on acoustic data, and determine the morphology of each underwater foreign object based on the distribution location of multiple underwater foreign objects and the corresponding image data; Step S13: Determine each contact risk node based on the morphology of the underwater part of the ship, optimize the ship's navigation attitude according to the node position of each contact risk node and the distribution position of each underwater foreign object, and determine the type of each underwater foreign object according to the morphology and corresponding color parameters. Step S14: Determine underwater foreign object events based on the actual number, distribution location, and type of each underwater foreign object; determine ship navigation events based on underwater foreign object events, water surface flow parameters, and ship's surrounding environmental parameters. Step S15: Determine multiple navigation risk factors for the ship based on the ship's navigation events and past navigation records, and determine the ship's handling method for underwater foreign objects based on the multiple navigation risk factors, underwater foreign object events, and the ship's navigation mode.

[0011] refer to Figure 2 In step S11, the specific steps are as follows: S111: The ship is in a navigation state and is sailing relative to the water surface. At this time, the underwater part of the ship is in an underwater scene; collect the ship's navigation direction and multiple navigation parameters, and determine the ship's navigation mode based on the ship's navigation direction, multiple navigation parameters and draft. S112: Based on the detection of the underwater part of the ship, determine multiple functional components of the underwater part, determine the corresponding safe range for underwater navigation according to the position, shape and function of multiple functional components, determine the underwater control space of the ship based on the safe range for underwater navigation, the ship's navigation mode and the ship's draft, and mark the morphological information of the underwater control space. At the same time, the underwater control space is dynamically adjusted with the dynamic navigation of the ship.

[0012] In the embodiments of this application, GPS and an inertial measurement unit (IMU) are used to monitor in real time whether the ship is in a navigation state (speed > 0.5 knots) and to confirm the relative position of the ship to the water surface; a pressure sensor is used to detect the contact state between the hull and the water surface to ensure that the ship is indeed "navigating relative to the water surface" rather than being grounded or suspended; when the ship's speed is stable and the pressure distribution of the hull is normal for 30 consecutive seconds, it is confirmed as a valid navigation state.

[0013] The system collects the ship's navigation direction and multiple navigation parameters, using an electric gyrocompass to obtain the ship's true heading (accuracy ±0.1°); speed parameters: measuring ground speed and water speed using a Doppler log; attitude parameters: measuring roll, pitch, and bow angles using an IMU; power parameters: main engine speed and propeller thrust; environmental parameters: wind speed, wind direction, current speed, and current direction.

[0014] The navigation mode of a ship is determined based on its direction of travel, multiple navigation parameters, and draft. A ship's navigation mode refers to the operational strategy adopted by the ship under specific navigation environments, mission objectives, and its own state, including speed control, rudder angle adjustment, power distribution, safety redundancy, and port maneuvering. The determination of the navigation mode is mainly based on three core inputs: direction of travel: the deviation between the ship's current course and the planned course, turning requirements, etc.; multiple navigation parameters: including but not limited to speed, rudder angle, propeller power, wind speed, current speed, and visibility; and draft: reflecting the ship's current load condition and directly affecting its maneuverability, stability, and ability to avoid underwater obstacles.

[0015] Furthermore, a high-resolution sonar scanning system (frequency 200-400kHz) is used to scan the underwater parts of the ship; the contour features of different functional components are identified through point cloud data processing algorithms; a three-dimensional model library of underwater components is established, including standard parameter templates; a basic safety distance matrix is ​​defined for each functional component; the safety factor is adjusted according to the functional importance of the component; the vulnerability of the component and the severity of the consequences of damage are considered; the safety range calculation method is: basic safety distance = component size × safety factor; functional weight coefficients: propulsion component (engine): 1.5; control component (rudder): 1.3; structural component (hull, keel): 1.0; the dynamic adjustment factor is adjusted in real time according to the speed and sea state.

[0016] The safety ranges of each component are spatially integrated; the size and shape of the overall control space are adjusted according to the navigation mode; the influence of draft on the vertical control range is considered; the convex hull algorithm is used to merge the safety ranges of each component into the overall control space; navigation mode adjustment: high-speed mode: expand forward by 50% and to both sides by 20%; maneuvering mode: expand in all directions by 30%; berthing mode: expand forward by 20% and to both sides by 50%; the influence of draft: the lower boundary of the control space = draft + 1.5 meters safety margin.

[0017] Simultaneously, a polyhedral model is used to represent the control space; key node coordinates and patch information are recorded; a dynamic update mechanism is established; morphological information is marked: spatial boundary points: key coordinate points of the outer contour of the control space are recorded; patch information: normal vector, area, and type (front, rear, left, right, bottom) of each patch; real-time update frequency: updated once per second; the underwater control space is dynamically adjusted with the dynamic navigation of the ship, at which time a ship motion state prediction model is established; changes in environmental parameters are monitored in real time; Kalman filtering is used for smoothing adjustment; dynamic adjustment mechanism: heading change: the control space rotates in real time with the heading; speed change: the size of the control space is adjusted according to the speed; roll / pitch: the control space tilts with the ship's attitude.

[0018] refer to Figure 3 In step S12, the specific steps are as follows: S121: Detect the underwater control space, determine the detection data set based on the detection of the underwater control space, determine the acoustic data and image data according to the screening of the detection data set, and mark the detection position of the acoustic data and the detection position of the image data; S122: For acoustic data, determine a preliminary shape map of the underwater control space based on the identification of the acoustic data, determine multiple features to be detected based on the preliminary shape map of the underwater control space and the shape of the underwater part, determine the corresponding underwater foreign objects based on the multiple features to be detected, the detection location of the acoustic data and the data content, and mark the distribution location of the multiple underwater foreign objects. S123: Based on the matching of the distribution location of multiple underwater foreign objects with the detection location of image data, the image part corresponding to each underwater foreign object is determined, and multiple morphological features of the underwater foreign object are determined based on the identification of the image part corresponding to each underwater foreign object, so as to determine the morphology of each underwater foreign object.

[0019] In the embodiments of this application, the underwater control space is detected to obtain acoustic and image information of the target water area and construct a basic data pool; acoustic scanning is performed using sonar; image acquisition is performed using an underwater camera or ROV (remotely operated underwater vehicle) equipped with a camera; sonar frequency: usually 100–500kHz, the higher the frequency, the better the resolution, but the shorter the detection distance; image resolution: it is recommended to be no less than 1080p to support subsequent foreign object identification.

[0020] The raw data undergoes initial cleaning and classification to construct a structured detection dataset; duplicate data, points with weak signals, or excessive location drift are removed; data is grouped by data type (sound wave / image) and timestamp; interpolation algorithms are used to fill data gaps caused by device jitter or signal loss; data cleaning is based on signal-to-noise ratio (SNR) thresholds, such as removing data points with SNR < 10dB; data alignment uses a timestamp + nearest neighbor method to ensure spatiotemporal consistency between sound wave and image data; simultaneously, sound wave and image data suitable for foreign object identification are extracted from the dataset and their precise locations are marked; sound wave data labeling records the latitude, longitude, depth, and reflection intensity of each sound wave point; image data labeling records the shooting location, shooting angle, and coverage area of ​​each image; a spatial index is constructed to facilitate subsequent querying and matching.

[0021] Specifically, the vessel conducts underwater foreign object inspection, defining the control area as a rectangular zone of 300m × 200m, with a depth range of 0–20m. A multibeam sonar is used for full-coverage scanning at a frequency of 300kHz, with the vessel traveling at 4 knots and its course parallel to the port's main channel. Simultaneously, the ROV, equipped with a 4K underwater camera, captures images along a zigzag path, taking one image every 10m. The positioning system records the latitude, longitude, and depth of each sonar ping and each image. Output: Sonar data: approximately 12,000 sound wave reflection points; Image data: approximately 150 underwater images; Location data: each data point includes latitude, longitude, and depth information.

[0022] Simultaneously, acoustic data points with an SNR < 10dB were removed (approximately 5% were filtered); blurry images were removed (such as low-contrast images caused by turbidity, approximately 10 images were filtered); acoustic data were interpolated to fill in a small number of data gaps caused by equipment rotation; a dataset was constructed in chronological order, with acoustic data and image data stored separately; acoustic dataset: 11,400 valid points; image dataset: 140 valid images; dataset format: JSON / CSV, including timestamps, latitude and longitude, depth, and data type.

[0023] Acoustic data processing: Each acoustic point is labeled with (longitude, latitude, depth, reflection intensity); Example point: (121.5123, 31.2345, -12.3, 45dB); Image data processing: Each image is labeled with (longitude, latitude, depth, pitch angle, yaw angle, coverage radius); Example image: (121.5125, 31.2347, -12.5, -30°, 45°, 5m); Spatial index construction: Using an R-tree index structure to support fast spatial queries; Output: Acoustic data: 11,400 labeled points; Image data: 140 labeled images; Spatial index file: Supports subsequent efficient queries.

[0024] Furthermore, the acoustic wave data is converted into a visualized underwater spatial morphology map, providing a foundation for subsequent foreign object identification; the acoustic wave data is preprocessed to convert the acoustic wave reflection data into a three-dimensional point cloud; interpolation algorithms (such as Kriging interpolation and inverse distance weighting) are used to construct a continuous seabed / riverbed surface model; the point cloud and terrain model are fused to generate a three-dimensional morphology map of the underwater space; optional point cloud density: no less than 10 points per square meter; interpolation accuracy: error controlled within ±5cm; morphology map resolution: it is recommended to have a grid of no less than 1m×1m.

[0025] Extract feature points or regions that may represent foreign objects from the morphological map; identify unnatural features in the terrain, such as protrusions, depressions, and linear structures; use edge detection, region growing, and clustering algorithms to identify potential foreign objects; calculate the size, shape, orientation, and reflectivity of each feature; feature size threshold: usually set to 0.5m × 0.5m or larger; shape complexity: measured using indicators such as perimeter-area ratio; reflectivity anomaly threshold: ±10dB relative to the surrounding environment.

[0026] Based on the features to be detected, the location of the acoustic wave data, and the data content, underwater foreign objects are identified and their distribution locations are marked. The system confirms whether the features to be detected are indeed foreign objects and pinpoints their location. Simultaneously, machine learning algorithms (such as random forests and SVMs) or rule bases are used to determine whether a feature is a foreign object. Combining the precise location information from the acoustic wave data, the center coordinates of the foreign object are calculated. A recognition confidence level is given based on feature prominence, data quality, etc. The recognition confidence threshold is typically set above 70%. Location accuracy is controlled within ±0.5m. Foreign object classification includes objects such as shipwreck debris, rocks, and artificial obstacles.

[0027] Specifically, the process involves: acquiring acoustic data using multibeam sonar; removing outliers (such as noise points with reflection intensity > 60 dB); applying sound velocity profile correction to eliminate propagation errors in the water; converting the processed acoustic data into a point cloud containing 11,400 points; a point cloud example: (121.5123, 31.2345, -12.3, 45 dB); using inverse distance weighting for interpolation to generate a 1m × 1m grid of seabed topography; the topographic map shows the channel center depth as -15m, gradually decreasing to -8m on both sides; fusing the point cloud and topographic data to generate a 3D morphological map; the map clearly shows the channel topography, artificial structures (such as breakwaters), and potential anomalies; output: a preliminary morphological map of the underwater control space (3D visualization model); and a topographic data file (containing elevation information for each grid point).

[0028] Morphological Analysis: Five abnormal regions were identified in the morphological map; Region 1: Circular protrusion, approximately 2m in diameter; Region 2: Linear structure, approximately 5m long and 0.5m wide; Region 3: Irregular depression, approximately 3m² in area; Feature Extraction: Feature boundaries were extracted using the Canny edge detection algorithm; Adjacent feature points were grouped using the DBSCAN clustering algorithm; Feature Quantization: Region 1: Circularity 0.85, average reflectivity 55dB; Region 2: Aspect ratio 10:1, average reflectivity 50dB; Region 3: Depth variation 1.5m, average reflectivity 35dB; Output: A list of five features to be detected (including attributes such as location, size, shape, and reflectivity); Feature distribution map (feature locations are marked on the morphological map).

[0029] Five features were classified using a pre-trained random forest model. Model inputs included shape parameters, reflection intensity, and depth variations. The results were as follows: Feature 1: Shipwreck debris (92% confidence); Feature 2: Abandoned pipe (85% confidence); Feature 3: Natural rock (78% confidence); Features 4-5: Natural seabed undulations (non-foreign objects). Location refinement: Feature 1 center coordinates: (121.5128, 31.2349, -12.7); Feature 2 center coordinates: (12... 1.5132, 31.2352, -11.5); Feature 3 center coordinates: (121.5135, 31.2355, -13.2); Confidence assessment: comprehensively considering factors such as data quality and feature significance; finally, 3 foreign objects were identified, all with a confidence level >75%; Output: underwater foreign object list (including type, location, size, and confidence level); foreign object distribution map (precisely marking the location of foreign objects on the morphological map); identification report (detailed description of the characteristics and judgment basis of each foreign object).

[0030] Therefore, the image portion corresponding to each underwater foreign object is determined by matching the distribution location of multiple underwater foreign objects with the detection location of image data. Based on the identification of the image portion corresponding to each underwater foreign object, multiple morphological features of the underwater foreign object are determined to determine the morphology of each underwater foreign object. This method takes into account the overall consideration of identifying the image portion corresponding to each underwater foreign object, ensuring the accuracy of multiple morphological features of the underwater foreign object.

[0031] At this point, the location of the foreign object identified by the acoustic wave is spatially matched with the image data to find the image region corresponding to each foreign object using a nearest neighbor matching algorithm; the shooting range and angle of the underwater camera are considered; a transformation relationship between the acoustic coordinate system and the image coordinate system is established; matching accuracy is controlled by setting a position tolerance range (usually 0.5-1 meter); the influence of factors such as water depth and current on the position is considered; the matching result is verified by cross-validation through multi-view images; feature point matching technology is used to improve accuracy; at this point, the position tolerance is 0.8 meters; the matching confidence threshold is 80%; the maximum matching distance is 2 meters (any distance exceeding this distance is considered to be without a corresponding image); the morphological features of the foreign object are extracted from the matched images to form the morphological characteristics. Provide the basis for identification; remove blur and color cast from underwater images; enhance contrast and edge features; extract key points using SIFT (Scale Invariant Feature Transform) or SURF algorithms; extract shape features using HOG (Histogram of Oriented Gradients); extract high-level features using deep learning models (such as CNN); morphological feature calculation: geometric features: aspect ratio, roundness, complexity, etc.; texture features: roughness, uniformity, directionality, etc.; color features: hue distribution, saturation, etc. (applicable to visible light images); optional, feature extraction window size: 64×64 pixels; feature vector dimension: 128 dimensions (SIFT features); number of morphological features: 20-30 features extracted per object.

[0032] The morphology of each underwater foreign object is determined based on its morphological features. The extracted morphological features are comprehensively analyzed to ultimately determine the specific morphology of each object. Support Vector Machine (SVM) or Random Forest is used for classification. Deep learning models are applied for end-to-end morphological recognition. Morphological description generation: A structured description is generated based on morphological features, including information such as shape, size, material, and state. The reliability of feature extraction is comprehensively considered. The credibility of the classification results is evaluated. Classification accuracy requirement: >90%. Morphological description dimensions: shape, size, material, state, completeness, etc. Confidence calculation weights: feature quality 40%, classification result 60%.

[0033] Specifically, S122 has identified three foreign objects, and the corresponding image data needs to be matched; Foreign object 1 (shipwreck): (121.5128, 31.2349, -12.7); Foreign object 2 (abandoned pipe): (121.5132, 31.2352, -11.5); Foreign object 3 (natural rock): (121.5135, 31.2355, -13.2); Image location data: Image group 1: Shooting range (121.5125-121.5130, 31.2). Image group 1: 345-31.2350); Image group 2: Shooting range (121.5130-121.5135, 31.2350-31.2355); Image group 3: Shooting range (121.5135-121.5140, 31.2355-31.2360); Matching execution: Foreign object 1 → Image group 1 (distance 0.3 meters, matching successful); Foreign object 2 → Image group 2 (distance 0.4 meters, matching successful); Foreign object 3 → Image group 2 (distance 0.6 meters, matching successful).

[0034] Verify matching accuracy by analyzing the overlapping areas of adjacent images; confirm that the confidence level of all matches is >85%; Output: Images corresponding to foreign object 1: IMG_0012.jpg to IMG_0018.jpg (7 images); Images corresponding to foreign object 2: IMG_0025.jpg to IMG_0032.jpg (8 images); Images corresponding to foreign object 3: IMG_0033.jpg to IMG_0040.jpg (8 images).

[0035] Feature extraction was performed on the matched images; IMG_0012.jpg to IMG_0040.jpg were deblurred; histogram equalization was used to enhance contrast; adaptive filtering was applied to remove noise; Feature extraction: Foreign object 1 image group: 245 SIFT feature points were extracted; the aspect ratio was calculated to be 2.3, roundness to be 0.35, and complexity to be 0.78; texture analysis showed metal corrosion features; Foreign object 2 image group: 189 SIFT feature points were extracted; the aspect ratio was calculated to be 8.5, roundness to be 0.12, and complexity to be 0.92; texture analysis showed regular tubular structures; Foreign object 3 image group: 312 SIFT feature points were extracted; the aspect ratio was calculated to be 1.2, roundness to be 0.68, and complexity to be 0.45; texture analysis showed natural rock textures.

[0036] Verify consistency of multiple image features; remove abnormal feature points (such as false features caused by reflected light spots); output: Morphological features of foreign object 1: metal surface, irregular shape, with corrosion marks; Morphological features of foreign object 2: tubular structure, smooth surface, about 5 meters in length; Morphological features of foreign object 3: blocky structure, rough surface, natural texture.

[0037] Foreign object morphology was determined based on morphological features, and morphological classification was performed as follows: Foreign object 1: Input features: metallic surface, irregular shape, corrosion marks; SVM classification result: shipwreck debris (confidence 94%); Deep learning validation: shipwreck debris (confidence 92%); Foreign object 2: Input features: tubular structure, smooth surface, 5 meters in length; SVM classification result: abandoned pipe (confidence 89%); Deep learning validation: abandoned pipe (confidence 91%); Foreign object 3: Input features: blocky structure, rough surface, natural texture; SVM classification result: natural rock (confidence 96%); Deep learning validation: natural rock (confidence 95%).

[0038] Morphological descriptions generated: Foreign object 1: Shape: Irregular block; Size: Approximately 3m × 2m × 1.5m; Material: Metal (suspected steel); Status: Severely corroded, structurally damaged; Foreign object 2: Shape: Long straight tube; Size: Approximately 5 meters long and 0.5 meters in diameter; Material: Metal (suspected cast iron); Status: Surface intact, open at both ends; Foreign object 3: Shape: Nearly elliptical block; Size: Approximately 2m × 1.8m × 1.2m; Material: Natural rock; Status: Intact, with marine life attached to the surface.

[0039] Confidence Assessment: Foreign Object 1: Overall confidence 93%; Foreign Object 2: Overall confidence 90%; Foreign Object 3: Overall confidence 95.5%; Output: Final form of Foreign Object 1: Shipwreck debris, metallic material, severely corroded; Final form of Foreign Object 2: Abandoned pipe, cast iron material, structurally intact; Final form of Foreign Object 3: Natural rock, with marine organisms attached to the surface; Morphology Recognition Report: Includes detailed description, confidence assessment, and recommended handling methods.

[0040] refer to Figure 4 In step S13, the specific steps are as follows: S131: Collect the morphology of the underwater part of the ship and mark the shape and function of multiple functional components of the underwater part. Based on the shape, function and navigation direction of the multiple functional components, determine the contact nodes of each functional component. Based on the contact nodes of each functional component and the corresponding risk coefficient, determine the contact risk nodes of the ship. S132: Collect the node positions of each contact risk node, determine the underwater foreign object risk distribution map based on the node positions of each contact risk node and the distribution positions of each underwater foreign object, determine the contact priority of each contact risk node based on the identification of the underwater foreign object risk distribution map, and trigger the optimization of the ship's navigation attitude based on the contact priority of each contact risk node, the ship's navigation attitude and the ship's navigation direction to avoid the impact of underwater foreign objects. S133: Collect the morphology of each underwater foreign object, determine the current color image of each underwater foreign object based on the detection of its distribution location, determine the color parameters of each underwater foreign object based on the recognition of its current color image, and determine the type of each underwater foreign object based on its morphology, corresponding color parameters, and type mapping relationship.

[0041] In the embodiments of this application, the morphology of the underwater portion of the ship is acquired, and the underwater portion of the hull is scanned using a multibeam sonar system or an underwater laser scanner; high-precision point cloud data is generated and registered into the ship's coordinate system; the output is a three-dimensional digital model of the underwater portion of the ship; based on the design drawings and scanning results, functional components are identified, such as: propeller (propulsion); rudder (steering); bottom plate (support); bulbous bow (drag reduction); depth sounder (measurement); for each component, the following are recorded: shape parameters: size, shape, structural complexity; function type: propulsion, steering, support, measurement, etc.; spatial position: three-dimensional coordinates relative to the ship's coordinate system; a functional component schematic table is acquired, as shown in Table 1. Table 1: Functional Component Diagram Based on the shape, function, and current navigation direction of the components, determine the contact node of each functional component. The contact node is the spatial point on the functional component that is most likely to first come into contact with the underwater foreign object. It is usually located at the front end or outermost edge of the component and is related to the navigation direction.

[0042] Obtain the ship's current heading angle (e.g., 45°); calculate the navigation direction vector (e.g., (1,1,0)); calculate the foremost point of each component in the navigation direction; for example, the blade tip of the propeller, the leading edge of the rudder, and the foremost point of the bulbous bow; the contact node of propulsion components (e.g., propeller) is usually at the blade tip; the contact node of steering components (e.g., rudder) is at the leading edge; the contact node of support components (e.g., bottom plate) is at the leading edge edge; collect a contact node diagram table, as shown in Table 2: Table 2. Schematic Diagram of Contact Nodes By combining the location of the contact points with the risk coefficient, high-risk contact points are selected. The risk coefficient reflects the degree of impact on ship safety after the component is damaged. It is usually based on: the importance of the component (e.g., a damaged propeller will result in loss of propulsion); the difficulty of repair (e.g., a damaged bulbous bow requires dry dock repair); and the consequences of the damage (e.g., a damaged rudder will result in loss of control).

[0043] Risk coefficient (1-10 points, 10 being the highest risk): Propeller: 9 points; Rudder: 8 points; Bulbous bow: 6 points; Bottom plate: 5 points; Depth sounder: 3 points; Contact risk node determination: Set a risk threshold (e.g., ≥7 points is a high-risk node); Screen out contact nodes with a risk coefficient ≥ the threshold; Collect a risk node illustration table, as shown in Table 3: Table 3. Risk Node Diagram Furthermore, the node locations of each contact risk node are collected, and an underwater foreign object risk distribution map is determined based on the node locations of each contact risk node and the distribution locations of each underwater foreign object. The contact priority of each contact risk node is determined based on the identification of the underwater foreign object risk distribution map. The ship's navigation attitude is optimized based on the contact priority of each contact risk node, the ship's navigation attitude, and the ship's navigation direction to avoid the impact of underwater foreign objects. This approach incorporates the overall considerations for the identification of the underwater foreign object risk distribution map and ensures the accuracy of the contact priority of each contact risk node.

[0044] At this point, the node positions of each contact risk node are collected from the risk node schematic table output by step S131; the node positions are represented in the ship coordinate system, usually in the format (x,y,z), and the unit is meters; the ship coordinate system is converted to a geographic coordinate system (such as WGS84) to facilitate spatial matching with the underwater foreign object position; the coordinate transformation is performed using the ship's real-time position (GPS) and attitude (heading angle, pitch angle, roll angle).

[0045] Construct a spatial risk distribution map to indicate the spatial relationship between each risk node and underwater foreign objects; perform spatial overlay analysis of the geographic coordinates of the risk nodes and the distribution location of the underwater foreign objects; calculate the Euclidean distance from each risk node to each underwater foreign object; risk assessment: set risk levels according to distance: distance < 5m: high risk (red); 5m ≤ distance < 10m: medium risk (yellow); distance ≥ 10m: low risk (green); risk distribution map construction: generate risk distribution maps using a GIS system or 3D visualization tools; the map includes: risk node location; underwater foreign object location; risk level areas (red, yellow, green).

[0046] Based on risk level, distance, and component importance, the contact priority of each risk node is determined; Priority assessment model: Priority = Risk level weight × Distance weight × Component importance weight; Risk level weight: High risk = 3, Medium risk = 2, Low risk = 1; Distance weight: 1 / Distance (the smaller the distance, the greater the weight); Component importance weight: Propeller = 1.5, Rudder = 1.2, Others = 1.0; Priority ranking: Calculate the priority score for each risk node; Sort by score from high to low.

[0047] Meanwhile, based on the priority ranking results, the ship's navigation attitude is dynamically adjusted to avoid high-risk encounters; for high-priority risk nodes, the course or depth is adjusted first; the adjustment methods include: changing the course angle: avoiding the foreign object area; adjusting the draft: surfacing or submerging to avoid foreign objects; reducing the speed: increasing reaction time; control system triggering: the ship's automatic control system receives optimization instructions; executes attitude adjustment, and monitors the adjustment effect in real time.

[0048] Specifically, a ship is navigating in Channel A. Step S131 identifies two contact risk nodes: N01: propeller, ship coordinate system position (0.5, -2.5, -5); N02: rudder, ship coordinate system position (0.2, 3.2, -4). At the same time, sonar detects two underwater foreign objects: W01: shipwreck debris, geographic coordinates (22.295°N, 114.175°E, -5.2); W02: abandoned container, geographic coordinates (22.295°N, 114.175°E, -4.8).

[0049] Convert N01 and N02 to geographic coordinates; Construct risk distribution map: Calculate distances: N01 to W01: 3.2m (high risk); N01 to W02: 8.7m (medium risk); N02 to W01: 4.8m (high risk); N02 to W02: 7.6m (medium risk); Determine contact priority: Calculate and sort priority scores; Optimize navigation attitude: For the highest priority N01-W01 combination, take measures to yaw the course 5° to the right and reduce the speed to 8 knots; For the second priority N02-W01 combination, take measures to increase the draft by 0.5m.

[0050] Therefore, the morphology of each underwater foreign object is collected, and the current color image of each underwater foreign object is determined based on the detection of its distribution location. The color parameters of each underwater foreign object are determined based on the recognition of its current color image. The type of each underwater foreign object is determined based on the mapping relationship between its morphology, corresponding color parameters, and type. This approach takes into account the overall consideration of the morphology, corresponding color parameters, and type mapping relationship of each underwater foreign object, ensuring the accuracy of the identification of each underwater foreign object type. At the same time, the introduction of the morphology of the underwater foreign object, taking into account the overall consideration of the morphology and corresponding color parameters of each underwater foreign object, improves the accuracy of the identification of each underwater foreign object type.

[0051] At this stage, the morphology of each underwater foreign object is collected, and an underwater high-definition camera system (such as a 4K resolution underwater camera) is used to photograph the foreign objects. The camera is usually equipped with LED supplementary lights to ensure clear images even in turbid water. A multi-angle shooting strategy is adopted, with at least 3-5 images of each foreign object taken from different angles. Position matching: The camera position is precisely matched with the distribution position of the foreign objects. An underwater positioning system (such as an ultra-short baseline positioning system) is used to ensure that the camera is facing the foreign objects. The shooting angle, distance, and lighting conditions of each image are recorded. Image preprocessing: Color correction is performed to compensate for the absorption and scattering effects of light by the water. Image enhancement algorithms are used to improve contrast and clarity. The multi-angle images are registered and fused to generate a complete surface texture map of the foreign objects.

[0052] Extract color feature parameters of foreign objects from the image; convert the RGB image to the HSV color space for easier color analysis; analyze the three components of hue (H), saturation (S), and lightness (V) respectively; color feature extraction: calculate the dominant colors: use the K-means clustering algorithm to cluster the image pixels by color; extract the 3-5 colors with the largest proportion as the dominant colors; calculate color distribution features: color uniformity: calculate the standard deviation of the color distribution; color contrast: calculate the color difference between adjacent areas; color texture features: analyze the color texture using the gray-level co-occurrence matrix; color parameter quantization: convert color features into numerical parameters: dominant hue: represented by the H value in the HSV space (0-360°); mean saturation: 0-100%; mean lightness: 0-100%; color complexity: the number of color types and distribution entropy.

[0053] Foreign object types are determined based on morphological and color parameters and type mapping relationships. Morphological features (volume, surface area, aspect ratio, etc.) and color features (hue, saturation, etc.) are combined into feature vectors. Principal component analysis (PCA) is used for feature dimensionality reduction to extract the most discriminative features. A database of foreign object types is established, containing typical features of each type: metals: high density, regular shape, metallic luster; rocks: irregular shape, natural texture, low saturation; wood: medium density, fibrous texture, brownish-yellow tone; plastics: low density, bright color, high saturation. Machine learning classifiers (such as SVM or random forest) are used for type identification. Deep learning methods (such as CNN) are used to learn features directly from the original images and morphological data. Confidence assessment: The classification results are assessed for confidence. When the confidence is below a threshold, it is marked as "unknown" and manual review is recommended.

[0054] Specifically, when a ship was navigating in port A, its sonar system detected two underwater foreign objects. It was necessary to determine their type to assess the risk. Multibeam sonar was used to scan the two objects: W01: 6.2m long, 2.5m wide, and 2.0m high, exhibiting hull structural characteristics; W02: 2.4m long, 2.4m wide, and 2.6m high, a standard cubic structure. Color image acquisition: An underwater robot was dispatched to take pictures: W01: A metal structure covered with reddish-orange rust was found, with a clear hull outline; W02: A regular cube painted in blue-gray was found, exhibiting characteristics of container corner fittings.

[0055] Color parameter extraction: W01: Dominant hue: 35° (red-orange); Saturation: 75%; Brightness: 82%; Color complexity: 0.65 (uneven rust distribution); W02: Dominant hue: 210° (blue); Saturation: 45%; Brightness: 60%; Color complexity: 0.42 (uniform color); Category identification: Feature fusion analysis: W01: Large metal structure + hull shape + rust characteristics → shipwreck (confidence 92%); W02: Standard cube + container corner pieces + blue-gray paint → abandoned container (confidence 88%); The ship adjusted its course based on the identification results, successfully avoiding the shipwreck and maintaining a safe distance from the abandoned container, ensuring navigational safety; The maritime department marked the shipwreck based on the report and plans to salvage and clean it up later.

[0056] refer to Figure 5 In step S14, the specific steps are as follows: S141: Collect an underwater foreign object risk distribution map, determine the actual number of each underwater foreign object based on the identification of the underwater foreign object risk distribution map, and determine the first event coefficient based on the actual number of each underwater foreign object and its corresponding distribution location. S142: Determine the second event coefficient based on the actual quantity and corresponding type of each underwater foreign object, and determine the underwater foreign object event based on the mapping relationship between the first event coefficient, the second event coefficient and the abnormal event; the underwater foreign object event presents the triggering factors of each underwater foreign object and reflects the severity of the underwater environment; S143: Collect water surface flow parameters and determine the ship's surrounding environmental parameters based on the ship's environmental monitoring. Determine the ship's first navigation obstruction factor based on underwater foreign object events and water surface flow parameters. Determine the ship's second navigation obstruction factor based on underwater foreign object events and the ship's surrounding environmental parameters. Determine the ship's navigation event based on the mapping relationship between the first navigation obstruction factor, the second navigation obstruction factor, and the navigation event.

[0057] In the embodiments of this application, an underwater foreign object risk distribution map is collected, the actual number of each underwater foreign object is determined based on the identification of the underwater foreign object risk distribution map, and a first event coefficient is determined based on the actual number of each underwater foreign object and its corresponding distribution location. This approach takes into account the overall consideration of the actual number of each underwater foreign object and its corresponding distribution location, ensuring the accuracy of the first event coefficient.

[0058] At this point, obtain structured layer data containing the spatial distribution, risk level, and impact range of underwater foreign objects; data source: "Underwater Foreign Object Risk Distribution Map" output from previous steps (such as S132); the map includes: coordinates (longitude, latitude, depth) of each foreign object; risk level (such as high, medium, low); and impact radius (such as 10m, 20m, 30m).

[0059] Contact priority (levels 1-5); Technical means: Use GIS (Geographic Information System) to perform spatial overlay analysis on the risk map; perform cluster analysis on foreign object points to identify high-density areas; mark overlapping influence areas and calculate the foreign object influence density.

[0060] The actual number of valid underwater foreign objects is calculated from the risk distribution map, and duplicate or mislabeled points are removed; foreign object points with similar coordinates (e.g., distance less than 5m) are merged; the DBSCAN clustering algorithm is used to identify and merge cases where the same foreign object is marked multiple times; all foreign object points are traversed and uniquely identified foreign objects are recorded; the total number of foreign objects in each area or the entire area is output.

[0061] The first event coefficient is determined based on the actual quantity and distribution location of the foreign objects. The "first event coefficient" (C1) is calculated using the quantity and spatial distribution density of the foreign objects to quantify the potential impact of foreign object events on navigation. The calculation formula is as follows: in: n : Number of foreign objects; di : No. i The closest distance between an object and the ship's planned route (unit: meters); wi : No. i Risk weights for individual foreign objects (e.g., high = 1.0, medium = 0.7, low = 0.5); the closer the object, the greater the impact; the higher the risk level, the greater the weight; the larger the coefficient, the higher the overall risk.

[0062] Specifically, the planned route for the vessel is from (120.50, 23.10) to (120.60, 23.20); a foreign object distribution diagram (after deduplication) is collected, as shown in Table 4. Table 4: Schematic Diagram of Foreign Object Distribution Weighting: High risk: w=1.0; Medium risk: w=0.7; Calculate the contribution of each foreign object: 001: 1 / 30×1.0≈0.0333; 002: 1 / 50×0.7≈0.0140; 003: 1 / 20×1.0=0.0500; Summarize: C 1 = 0.0333 + 0.0140 + 0.0500 = 0.0973; Normalization (optional): If a scaling factor (e.g., ×100) is used, then C 1 = 9.73; Result Interpretation; C 1 = 9.73 (after normalization); this value is higher than the preset threshold (e.g., 7.0), indicating that the underwater foreign object poses a significant threat to the ship's navigation; it can proceed to the subsequent event assessment process (e.g., S142).

[0063] Furthermore, the second event coefficient is determined based on the actual quantity and corresponding type of each underwater foreign object, and the underwater foreign object event is determined based on the mapping relationship between the first event coefficient, the second event coefficient, and the abnormal event. The underwater foreign object event presents the triggering factors of each underwater foreign object and reflects the severity of the underwater environment. It takes into account the overall consideration of the first event coefficient, the second event coefficient, and the mapping relationship of the abnormal event, ensuring the accuracy of the underwater foreign object event.

[0064] At this point, the second event coefficient is determined based on the actual quantity and type of underwater foreign objects, quantifying the risk posed by the type of foreign object, and the second event coefficient is calculated in conjunction with the quantity. C 2) Definition of foreign objects: The types include, but are not limited to: shipwreck debris; fishing nets; rocks; metal tanks; abandoned pipelines; artificial obstacles (such as caissons and anchor chains).

[0065] Risk weighting by type: Shipwreck debris has a risk weight of 1.0; fishing nets have a risk weight of 0.6; rocks have a risk weight of 0.8; metal tanks have a risk weight of 0.5; abandoned pipelines have a risk weight of 0.4; artificial obstacles have a risk weight of 0.7; calculation examples are shown in Table 5. Table 5 Calculation Example Table Calculation process: C 2 = 2.0 + 1.8 + 0.8 + 2.0 + 0.8 + 0.7 = 8.1; Underwater foreign object events are determined based on the first event coefficient, the second event coefficient, and the mapping relationship of abnormal events, comprehensively... C 1 (first event coefficient) and C 2 (Second Event Coefficient): The type of underwater foreign object event is determined through mapping relationships. An event mapping table is introduced, as shown in Table 6: Table 6 Event Mapping Table Comprehensive judgment logic: If C 1 and C If both fall within the same interval, a direct judgment is made; if they are not in the same interval, the higher risk factor is used. Assumption: C 1 = 9.73 (from S141); C 2 = 8.1; Judgment process: C 1∈[5,10], C 2∈[5,10]→Medium-risk event; Result: Event type: Medium-risk event; Triggering factors: Distribution of various foreign objects (shipwrecks, fishing nets, rocks, etc.); Severity: Medium.

[0066] Therefore, by collecting surface flow parameters and determining the ship's surrounding environmental parameters based on environmental monitoring, the first navigation obstruction factor is determined based on underwater foreign object events and surface flow parameters, and the second navigation obstruction factor is determined based on underwater foreign object events and surrounding environmental parameters. The ship's navigation events are then determined based on the mapping relationship between the first and second navigation obstruction factors and navigation events. This comprehensive approach, which considers the mapping relationship between the first and second navigation obstruction factors and navigation events, ensures the accuracy of ship navigation events.

[0067] At this point, relevant parameters of water surface flow are acquired to assess the impact of water flow on ship navigation; data sources include shipborne current meters, buoy monitoring stations, satellite remote sensing, and meteorological and hydrological stations; ensure that the data is synchronized with the ship's position; if the data points are sparse, methods such as Kriging interpolation are used to supplement them; outliers are removed to ensure data smoothness.

[0068] Acquire information about the ship's surrounding environment to assess navigation safety; data sources include shipborne radar, LiDAR, AIS (Automatic Identification System), cameras, and weather sensors; combine radar, AIS, and camera data for comprehensive analysis; and dynamically update the status of the ship's surrounding environment.

[0069] Based on underwater foreign object events and surface flow parameters, determine the first navigational obstruction factor and analyze the combined effect of underwater foreign objects and water flow on ship navigation; Input: underwater foreign object event (from S142); surface flow parameters; Analysis logic: superimposed analysis of foreign object location and water flow direction: if the foreign object is located downstream, the ship is easily pushed towards the foreign object; if the water flow velocity is high, the ship faces increased difficulty in avoiding it; combined analysis of wave height and foreign object depth: when the waves are high, the ship's undulation is large, making it easy to hit shallow foreign objects; Output: first navigational obstruction factor: type: such as "water flow-driven obstacle risk"; description: such as "high current velocity pushes the ship towards the wreckage";

[0070] Based on underwater foreign object (FOO) events and ship-surrounding environmental parameters, determine the second navigation obstruction factor and analyze the combined effect of underwater FEO and the ship's surrounding environment on navigation. Inputs: Underwater FEO event; ship-surrounding environmental parameters; Analysis logic: Visibility and FEO location combination: In low visibility, even knowing the FEO's location makes precise avoidance difficult; Surrounding ship density and avoidance space: When ships are densely packed, avoidance space is limited, increasing the risk of collision; Wind direction and ship maneuverability: Crosswinds affect the ship's turning ability, impacting avoidance efficiency; Output: Second navigation obstruction factor: Type: e.g., "Environmentally constrained avoidance difficulty"; Description: e.g., "Difficulty in accurately avoiding fishing nets in low visibility";

[0071] Based on the first and second navigation obstruction factors and the navigation event mapping relationship, the ship navigation events are determined. The navigation event mapping relationship is introduced and presented in the navigation event mapping relationship table, as shown in Table 7.

[0072] Table 7. Mapping Relationships of Navigation Events Specifically, the vessel is navigating in port A; underwater foreign object incident (S142 output): type: medium risk event; foreign object distribution: 2 shipwrecks, 3 fishing nets, 1 rock; water surface flow parameters: current speed: 2.5 knots; current direction: 120°; wave height: 1.2 meters; vessel surrounding environment parameters: visibility: 150 meters; surrounding vessel density: 5 vessels / square kilometer; nearest obstacle distance: 300 meters.

[0073] Identify the first obstacle to navigation: Analysis: High current speed (2.5 knots), current direction 120°, the ship is easily pushed towards the wreckage; wave height 1.2 meters, the ship's undulation is obvious, increasing the risk of contact with shallow fishing nets; Output: First obstacle: Risk of water-driven obstacle; Identify the second obstacle to navigation: Analysis: Visibility is only 150 meters, making it difficult to accurately avoid fishing nets; high density of surrounding ships, limiting avoidance space; Output: Second obstacle: Environmental constraint avoidance difficulty; Table lookup mapping: First obstacle: Water-driven obstacle; Second obstacle: Environmental constraint avoidance; Corresponding event type: High-risk avoidance event; 4. Output results: Ship navigation event: Event type: High-risk avoidance event; Risk level: High; Recommended operation: Immediately reduce speed, activate collision avoidance radar, adjust course to move away from the wreckage area.

[0074] refer to Figure 6 In step S15, the specific steps are as follows: S151: Based on the identification of ship navigation events, determine the main navigation items and multiple sub-navigation items, determine multiple risk characteristics based on the main navigation items, multiple sub-navigation items and previous navigation records, and determine multiple navigation risk factors of the ship based on the multiple risk characteristics, the service life of the ship and the underwater environment in which the ship is located. S152: Determine the impact level of underwater foreign objects based on multiple navigation risk factors and underwater foreign object events. At the same time, collect data on the ship's underwater foreign object handling equipment. Determine the ship's handling method for underwater foreign objects based on the impact level of the underwater foreign objects, the ship's underwater foreign object handling equipment, and the ship's navigation mode. At this time, different levels of handling events are carried out for different underwater foreign objects.

[0075] In the embodiments of this application, the main navigation items and multiple sub-navigation items are determined based on the identification of ship navigation events. Multiple risk characteristics are determined based on the main navigation items, multiple sub-navigation items and previous navigation records. Multiple navigation risk factors of the ship are determined based on the multiple risk characteristics, the service life of the ship and the underwater environment in which the ship is located. This approach takes into account the overall consideration of multiple risk characteristics, the service life of the ship and the underwater environment in which the ship is located, thus ensuring the accuracy of the multiple navigation risk factors of the ship.

[0076] At this point, the ship's current navigation events are analyzed into primary navigation items (core tasks) and secondary navigation items (auxiliary tasks). Primary navigation items represent the ship's current core navigation tasks, such as "normal navigation," "obstacle avoidance navigation," and "emergency stop." Secondary navigation items are specific operations performed to support the primary navigation items, such as "deceleration," "turning," and "activating detection equipment." The navigation events are automatically mapped to the primary navigation items based on their type (e.g., "high-risk avoidance events"). The primary navigation items are further decomposed into multiple secondary navigation items to ensure that each operation step is clear and executable. Output results: Primary navigation item: 1; Secondary navigation items: multiple (usually 3-5).

[0077] Based on historical navigation data, extract risk characteristics of the current navigation mission; Historical record analysis: retrieve navigation records similar to the current primary and secondary navigation projects from the historical database; extract risk events from historical records, such as equipment failure, collision, grounding, etc.; Risk feature extraction: Feature types: Operational risks: such as frequent turns, sudden stops, etc.; Environmental risks: such as severe weather, complex waterways, etc.; Equipment risks: such as propeller failure, navigation system failure, etc.; Feature quantification: convert risk features into quantifiable indicators, such as "operational risk index", "environmental risk index", etc.; Output results: Risk feature list: multiple (usually 5-10); each feature includes: feature name, risk level, historical occurrence frequency.

[0078] Multiple navigation risk factors for a vessel are determined based on various risk characteristics, the vessel's service life, and the underwater environment in which the vessel operates. The degree of equipment aging is assessed according to the vessel's service life: New vessels (0-5 years): Equipment in good condition, low risk; Middle-aged vessels (5-15 years): Some equipment is aging, medium risk; Old vessels (over 15 years): Equipment is severely aging, high risk. Underwater environment assessment: Environment type: Clean water: No obstacles, low risk; General water: Few obstacles, medium risk; Complex water: Many obstacles, high risk; Environmental parameters: Water depth, current velocity, visibility, obstacle density, etc.

[0079] Based on comprehensive risk characteristics, service life, and underwater environment, navigation risk factors are generated: equipment aging risk; environmental adaptability risk; obstacle avoidance operation risk; navigation accuracy risk; emergency response risk; output results: list of navigation risk factors: multiple (usually 5-8); each factor includes: factor name, risk level, and scope of impact.

[0080] Specifically, the vessel information is as follows: Vessel name: "No. A" vessel; Service life: 12 years (middle-aged vessel); Current navigation event: High-risk avoidance event (S143 output); Navigation environment: Water area: Channel A; Underwater environment: Complex water area (multiple shipwrecks, fishing nets, rocks); Water depth: 15 meters; Current speed: 2 knots; Visibility: Moderate.

[0081] Identify primary and secondary navigation events: Navigation events: High-risk avoidance events; Primary navigation event: Obstacle avoidance navigation; Secondary navigation event: Deceleration navigation; Activate sonar detection; Adjust course; Activate collision avoidance radar; Historical record retrieval: Similar events: 10 obstacle avoidance navigation records; Risk events: 3 equipment failures, 2 collision risks, 1 grounding risk; Risk feature extraction: Operational risks: Frequent turns (high risk); Environmental risks: Complex waterways (high risk); Equipment risks: Aging sonar equipment (medium risk); Navigation risks: Unstable GPS signals (medium risk); Emergency risks: Delayed emergency response (low risk).

[0082] Service life analysis: 12 years old: some equipment is aging (medium-high risk); Underwater environment assessment: Complex waters: many obstacles (high risk); Water depth 15 meters: moderate; Current speed 2 knots: moderate; Visibility moderate: medium risk; Risk factors: Equipment aging risk: medium-high; Environmental adaptability risk: high; Obstacle avoidance operation risk: high; Navigation accuracy risk: medium; Emergency response risk: low; Final output: Main navigation item: obstacle avoidance navigation; Secondary navigation item: deceleration navigation; activation of sonar detection; course adjustment; activation of collision avoidance radar; Risk characteristics: Operational risk: high; Environmental risk: high; Equipment risk: medium; Navigation risk: medium; Emergency risk: low; Navigation risk factors: Equipment aging risk: medium-high; Environmental adaptability risk: high; Obstacle avoidance operation risk: high; Navigation accuracy risk: medium; Emergency response risk: low.

[0083] Furthermore, the impact level of underwater foreign object (GMO) is determined based on multiple navigation risk factors and underwater GMO events. Simultaneously, data on the ship's GMO handling equipment is collected. Based on the GMO impact level, the ship's GMO handling equipment, and the ship's navigation mode, the ship's handling method for underwater GMOs is determined. At this point, different levels of handling events are applied to different underwater GMOs, incorporating a holistic consideration of the GMO impact level, the ship's GMO handling equipment, and the ship's navigation mode. This ensures the accuracy of the ship's GMO handling method. Additionally, the introduction of ship navigation events allows for a holistic consideration of multiple navigation risk factors, underwater GMO events, and the ship's navigation mode, further improving the accuracy of the ship's GMO handling method. The ship has thus achieved the detection and handling of underwater GMOs.

[0084] At this point, considering both navigation risk factors and underwater foreign object events, the impact level of the foreign object on the ship's navigation is determined; risk factor weight allocation: a weight is assigned to each navigation risk factor (e.g., equipment aging risk: 0.2, environmental adaptability risk: 0.3, obstacle avoidance operation risk: 0.3, navigation accuracy risk: 0.1, emergency response risk: 0.1); the total weight is 1 to ensure the objectivity of the assessment.

[0085] Foreign Object Incident Feature Extraction: Foreign Object Type: such as rocks, shipwrecks, fishing nets, floating objects, etc.; Foreign Object Location: distance and orientation relative to the ship; Foreign Object Size: length, width, height; Foreign Object Motion Status: fixed, moving, floating, etc.; Impact Level Calculation: A weighted scoring method is used to combine risk factors with foreign object characteristics to calculate the impact level; Impact levels are usually divided into: low (1-3 points), medium (4-6 points), and high (7-10 points); Output Results: Impact Level: low / medium / high; Scoring Details: contribution score of each risk factor.

[0086] Acquire information on currently available underwater foreign object handling equipment for the vessel; Equipment list collection: Robotic arm: for grasping and removing large foreign objects; Sonar equipment: for detecting and locating foreign objects; ROV (Remotely Operated Vehicle): for close-range observation and handling of foreign objects; High-pressure water jet: for removing foreign objects attached to the hull; Trawling equipment: for collecting floating and small foreign objects; Equipment status assessment: Equipment availability: check if the equipment is available; Equipment performance: assess the equipment's handling capacity (e.g., maximum grasping weight of the robotic arm, maximum working depth of the ROV); Equipment suitability: select the most suitable equipment based on the type of foreign object; Output results: List of available equipment: including equipment type, status, and performance parameters; Equipment suitability assessment: recommended equipment for the current foreign object.

[0087] The ship's handling method for underwater foreign objects (IGFOs) is determined based on the impact level of the IGFO, the ship's IGFO handling equipment, and the ship's navigation mode. The handling methods are mapped as follows: Low impact level: typically monitoring and minor adjustments; Medium impact level: requiring partial intervention with some handling equipment; High impact level: requiring full activation of handling equipment and potentially altering the navigation plan. Navigation mode considerations include: Normal navigation mode: prioritizing maintaining speed and course, with preventative handling; Obstacle avoidance navigation mode: prioritizing avoiding IGFOs, with proactive intervention; Emergency stopping mode: prioritizing ship safety, with emergency handling.

[0088] Handling Method Formulation: Equipment Selection: Select the most suitable equipment based on the impact level and equipment applicability; Operation Procedures: Develop detailed operation procedures, including equipment startup, operation sequence, safety measures, etc.; Emergency Plan: Develop emergency plans for high-impact levels to ensure rapid response in the event of handling failure; Output Results: Handling Method: Monitoring / Intervention / Emergency Handling; Equipment Selection: Specific equipment; Operation Procedures: Detailed operation procedures; Emergency Plan: Emergency response measures.

[0089] Specifically, a 15-year-old vessel was navigating in Channel A when its sonar detected a large rock (5m × 3m × 2m) 200 meters ahead. The vessel was currently in obstacle avoidance mode. Risk factor weights: Equipment aging risk: 0.2 (score: 7); Environmental adaptability risk: 0.3 (score: 8); Obstacle avoidance operation risk: 0.3 (score: 9); Navigation accuracy risk: 0.1 (score: 6); Emergency response risk: 0.1 (score: 4); Foreign object event characteristics: Type: Rock; Location: 200 meters ahead, in the center of the channel; Size: 5m × 3m × 2m; Motion status: Fixed; Impact level calculation: Weighted total score = 7 × 0.2 + 8 × 0.3 + 9 × 0.3 + 6 × 0.1 + 4 × 0.1 = 7.7; Impact level: High (7-10 points).

[0090] Available equipment: Robotic arm: Available, maximum gripping weight 10 tons; Sonar equipment: Available, detection range 500 meters; ROV (Remotely Operated Vehicle): Available, maximum working depth 100 meters; High-pressure water gun: Unavailable (under maintenance); Trawling equipment: Available, suitable for floating objects; Equipment suitability assessment: Recommended equipment for rock handling: Robotic arm, ROV.

[0091] Impact Level: High; Navigation Mode: Obstacle Avoidance Navigation; Handling Method: Equipment Selection: Robotic Arm, ROV; Operation Steps: Activate ROV for close-range observation; Use the robotic arm to attempt to remove the rock; If removal fails, adjust course to detour; Emergency Plan: If the robotic arm cannot remove the rock, immediately initiate the emergency stop procedure; Notify port management for assistance; Final Output: Impact Level: High (7.7 points); Available Equipment: Robotic Arm, Sonar, ROV, Trawler; Handling Method: Main Equipment: Robotic Arm, ROV; Operation Steps: ROV for close-range observation; Robotic arm to attempt removal; Adjust course if necessary; Emergency Plan: Emergency stop, request assistance.

[0092] Please see Figure 7 , Figure 7 This is a schematic diagram of the structural composition of a vision-based underwater foreign object detection system for ships, as described in this embodiment of the invention. The vision-based underwater foreign object detection system for ships includes: The underwater control space module 21 is used for the ship to navigate relative to the water surface and to determine the ship's underwater control space based on the ship's navigation direction, draft, and the shape of the ship's underwater parts. The multimodal data module 22 is used to determine image data based on visual detection of the underwater control space, determine the distribution location of multiple underwater foreign objects based on acoustic data, and determine the morphology of each underwater foreign object based on the distribution location of multiple underwater foreign objects and the corresponding image data. The classification module 23 is used to determine each contact risk node based on the shape of the underwater part of the ship, optimize the ship's navigation attitude according to the node position of each contact risk node and the distribution position of each underwater foreign object, and determine the type of each underwater foreign object according to the shape and corresponding color parameters. The ship navigation event module 24 is used to determine underwater foreign object events based on the actual number, distribution location and type of each underwater foreign object, and to determine ship navigation events based on underwater foreign object events, water surface flow parameters and ship surrounding environment parameters. The processing module 25 is used to determine multiple navigation risk factors of the ship based on the ship's navigation events and past navigation records, and to determine the ship's processing method for underwater foreign objects based on the multiple navigation risk factors, underwater foreign object events and the ship's navigation mode.

[0093] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

Claims

1. A visual inspection-based method for detecting underwater foreign objects by a ship, characterized in that, include: When a ship navigates relative to the water surface, its underwater control space is determined based on its navigation direction, draft, and the shape of its underwater components. Image data is determined based on visual inspection of the underwater control space; the distribution location of multiple underwater foreign objects is determined based on acoustic data; and the morphology of each underwater foreign object is determined based on the distribution location of multiple underwater foreign objects and the corresponding image data. Based on the morphology of the underwater part of the ship, each contact risk node is determined. The ship's navigation attitude is optimized according to the node position of each contact risk node and the distribution position of each underwater foreign object. The type of each underwater foreign object is determined according to the shape and corresponding color parameters of each underwater foreign object. Underwater foreign object events are determined based on the actual number, distribution location, and type of each underwater foreign object; ship navigation events are determined based on underwater foreign object events, water surface flow parameters, and ship's surrounding environmental parameters. Based on ship navigation events and past navigation records, multiple navigation risk factors for the ship are identified. Based on these multiple navigation risk factors, underwater foreign object incidents, and the ship's navigation patterns, the ship's handling methods for underwater foreign objects are determined.

2. The visual inspection-based method for detecting underwater foreign objects by ships according to claim 1, characterized in that, The vessel navigates relative to the water surface, and its underwater control space is determined based on the vessel's navigation direction, draft, and the shape of its underwater components, including: The ship is in a navigation state and is moving relative to the water surface. At this time, the underwater part of the ship is in an underwater scene; the ship's navigation direction and multiple navigation parameters are collected, and the ship's navigation mode is determined based on the ship's navigation direction, multiple navigation parameters and draft. Based on the detection of the underwater part of the ship, multiple functional components of the underwater part are identified. The corresponding safe range for underwater navigation is determined according to the location, shape and function of multiple functional components. The underwater control space of the ship is determined based on the safe range for underwater navigation, the ship's navigation mode and the ship's draft, and the morphological information of the underwater control space is marked. At the same time, the underwater control space is dynamically adjusted with the dynamic navigation of the ship.

3. The visual inspection-based method for detecting underwater foreign objects by ships according to claim 1, characterized in that, The process involves determining image data based on visual detection of the underwater control space, determining the distribution locations of multiple underwater foreign objects based on acoustic data, and determining the morphology of each underwater foreign object based on its distribution location and corresponding image data, including: The underwater control space is detected, and a detection data set is determined based on the detection of the underwater control space. Based on the screening of the detection data set, acoustic data and image data are determined, and the detection positions of acoustic data and image data are marked. Based on the acoustic data, a preliminary shape map of the underwater control space is determined. Based on the preliminary shape map of the underwater control space and the shape of the underwater part, multiple features to be detected are determined. Based on the multiple features to be detected, the detection location and data content of the acoustic data, the corresponding underwater foreign objects are determined, and the distribution location of multiple underwater foreign objects is marked. The image portion corresponding to each underwater foreign object is determined by matching the distribution location of multiple underwater foreign objects with the detection location of image data. Based on the identification of the image portion corresponding to each underwater foreign object, multiple morphological features of the underwater foreign object are determined to determine the morphology of each underwater foreign object.

4. The visual detection method for underwater foreign objects by ships according to claim 1, characterized in that, The process involves determining each contact risk node based on the morphology of the ship's underwater components, optimizing the ship's navigation attitude based on the node positions of each contact risk node and the distribution locations of each underwater foreign object, and determining the type of each underwater foreign object based on its shape and corresponding color parameters, including: The morphology of the underwater part of the ship is collected, and the shape and function of multiple functional components of the underwater part are marked. Based on the shape, function and navigation direction of the multiple functional components, the contact nodes of each functional component are determined. Based on the contact nodes of each functional component and the corresponding risk coefficient, the contact risk nodes of the ship are determined. The system collects the node locations of each contact risk node, determines the underwater foreign object risk distribution map based on the node locations of each contact risk node and the distribution locations of each underwater foreign object, determines the contact priority of each contact risk node based on the identification of the underwater foreign object risk distribution map, and triggers the optimization of the ship's navigation attitude based on the contact priority of each contact risk node, the ship's navigation attitude and the ship's navigation direction to avoid the impact of underwater foreign objects.

5. The visual detection method for underwater foreign objects by ships according to claim 4, characterized in that, The method of determining each contact risk node based on the morphology of the ship's underwater components, optimizing the ship's navigation attitude based on the node positions of each contact risk node and the distribution positions of each underwater foreign object, and determining the type of each underwater foreign object based on its morphology and corresponding color parameters, also includes: The morphology of each underwater foreign object is collected, and the current color image of each underwater foreign object is determined based on the detection of its distribution location. The color parameters of each underwater foreign object are determined based on the recognition of its current color image. The type of each underwater foreign object is determined based on its morphology, corresponding color parameters, and type mapping relationship.

6. The visual inspection-based method for detecting underwater foreign objects by a ship according to claim 1, characterized in that, The process of determining underwater foreign object events based on the actual quantity, distribution location, and type of each underwater foreign object, and determining ship navigation events based on underwater foreign object events, surface flow parameters, and surrounding environmental parameters of the ship, includes: Collect underwater foreign object risk distribution map, determine the actual number of each underwater foreign object based on the identification of the underwater foreign object risk distribution map, and determine the first event coefficient based on the actual number of each underwater foreign object and its corresponding distribution location; The second event coefficient is determined based on the actual number and corresponding type of each underwater foreign object. Underwater foreign object events are determined based on the mapping relationship between the first event coefficient, the second event coefficient, and the abnormal event. Underwater foreign object events present the initiating factors of each underwater foreign object and reflect the severity of the underwater environment.

7. The visual detection method for underwater foreign objects by ships according to claim 6, characterized in that, The method of determining underwater foreign object events based on the actual quantity, distribution location, and type of each underwater foreign object, and determining ship navigation events based on underwater foreign object events, surface flow parameters, and ship's surrounding environmental parameters, also includes: The system collects water surface flow parameters and determines the ship's surrounding environmental parameters based on environmental monitoring. It then determines the first navigation obstruction factor based on underwater foreign object events and water surface flow parameters, the second navigation obstruction factor based on underwater foreign object events and surrounding environmental parameters, and finally determines the ship's navigation events based on the mapping relationship between the first navigation obstruction factor, the second navigation obstruction factor, and navigation events.

8. The method for detecting underwater foreign objects by a ship based on visual inspection according to claim 1, characterized in that, The process of determining multiple navigation risk factors for a ship based on navigation events and past navigation records, and determining the ship's handling methods for underwater objects based on these multiple navigation risk factors, underwater foreign object incidents, and the ship's navigation patterns, includes: Based on the identification of ship navigation events, the main navigation items and multiple sub-navigation items are determined. Based on the main navigation items, multiple sub-navigation items and previous navigation records, multiple risk characteristics are determined. Based on the multiple risk characteristics, the service life of the ship and the underwater environment in which the ship is located, multiple navigation risk factors of the ship are determined.

9. The visual detection method for underwater foreign objects by a ship according to claim 8, characterized in that, The method of determining multiple navigation risk factors for a ship based on navigation events and past navigation records, and determining the ship's handling method for underwater foreign objects based on multiple navigation risk factors, underwater foreign object incidents, and the ship's navigation pattern, also includes: The impact level of underwater foreign object (GMO) is determined based on multiple navigation risk factors and underwater GMO events. At the same time, the ship's GMO handling equipment is collected. Based on the impact level of the GMO, the ship's GMO handling equipment, and the ship's navigation mode, the ship's GMO handling method is determined. At this point, different levels of handling events are carried out for different GMOs.

10. A vision-based detection system for underwater foreign objects on ships, characterized in that, The vision-based detection system for underwater foreign objects is applied to the vision-based detection method for underwater foreign objects as described in any one of claims 1-9, wherein the vision-based detection system for underwater foreign objects includes: The underwater control space module is used for the ship's navigation relative to the water surface. It determines the ship's underwater control space based on the ship's navigation direction, draft, and the shape of the ship's underwater parts. The morphology module is used to determine image data based on visual detection of the underwater control space, determine the distribution location of multiple underwater foreign objects based on acoustic data, and determine the morphology of each underwater foreign object based on the distribution location of multiple underwater foreign objects and the corresponding image data. The classification module is used to determine each contact risk node based on the shape of the underwater part of the ship, optimize the ship's navigation attitude according to the node position of each contact risk node and the distribution position of each underwater foreign object, and determine the type of each underwater foreign object according to the shape and corresponding color parameters. The ship navigation event module is used to determine underwater foreign object events based on the actual number, distribution location and type of each underwater foreign object, and to determine ship navigation events based on underwater foreign object events, water surface flow parameters and ship surrounding environment parameters. The handling method module is used to determine multiple navigation risk factors of a ship based on ship navigation events and past navigation records, and to determine the ship's handling method for underwater foreign objects based on multiple navigation risk factors, underwater foreign object events, and the ship's navigation mode.

Images