A method for ensuring fusion of new state and transient of marine geographic information of ship navigation

Abstract

The application discloses a kind of new state and transient fusion guarantee method of marine geographic information of ship navigation, belong to marine geographic information and navigation safety guarantee technical field.The present application includes new state and transient fusion guarantee mode.New state and transient fusion guarantee mode is based on the latest chart information that has been corrected, by remote control small unmanned ship carries high-resolution multi-beam detection device, real-time acquisition of local sea area in front of ship along planned route transient marine geographic information, automatically identify navigation hazard and generate safe obstacle avoidance path.The present application overcomes the defects of the existing new state product present situation deficiency and transient product limited coverage, realizes the navigation safety guarantee of full voyage, high present situation, especially suitable for unfamiliar waters, variable water area and other complex sea area of ship navigation.

Description

Technical Field

[0001] This invention belongs to the field of marine geographic information and ship navigation safety assurance technology, and relates to a method for ensuring safety based on the fusion of novel and transient marine geographic information. Background Technology

[0002] The International Maritime Organization (IMO) and national maritime safety authorities have clearly stipulated that uncorrected basic marine geographic information products (such as paper charts and electronic nautical charts (ENCs) that have not been corrected in a timely manner, hereinafter collectively referred to as "charts") cannot be used as legally valid navigation safety assurance products for ships, and can only be used as reference materials. New marine geographic information products, if corrected to the latest version as required, are legally valid navigation safety assurance products and can generally meet the needs of ship navigation safety. However, they still suffer from insufficient timeliness when navigating in unfamiliar waters that have not been measured or precisely measured, in volatile waters affected by extreme factors such as war, earthquakes, typhoons, and other complex waters. Transient marine geographic information products acquire real-time underwater geographic information of the surrounding local sea area through autonomous ship detection, possessing the characteristics of objectivity, accuracy, and reliability, but their effective coverage is very limited and cannot independently complete the entire voyage safety assurance mission. Summary of the Invention

[0003] This invention provides a method for ensuring navigation safety by fusing new and transient marine geographic information, overcoming the shortcomings of insufficient timeliness of existing new products and limited coverage of transient products, and achieving full-range, high timeliness, and high reliability navigation safety assurance.

[0004] A method for ensuring the fusion of novel and transient marine geographic information for ship navigation, including a novel and transient fusion assurance mode. It includes the following steps: (1) Based on new marine geographic information covering the entire navigation area, derived from updated nautical charts, the planned route is set. Figure 1 As shown, the base map on the chart is the latest corrected nautical chart. The red cross symbols on the chart represent navigational obstructions (reefs) added after the correction, and the yellow line represents a planned route from south to north set by the ship using this nautical chart before sailing.

[0005] (2) The ship sails along the planned route, and a remotely controlled small unmanned vessel equipped with a high-resolution multibeam detection device conducts real-time scanning and detection along the planned route in front of the ship to obtain the transient marine geographic information required to achieve integrated support. To obtain transient information that meets the requirements of fusion processing, the raw detection data needs to undergo noise reduction, filtering, correction, thinning, and rapid identification of underwater navigational obstructions. Specifically: 1) Noise Reduction and Filtering First, by combining statistical outlier removal, radius filtering, and adaptive filtering based on local terrain features, noise such as suspended matter in the water and detected anomalies is eliminated while strictly preserving the sparse multibeam point cloud corresponding to isolated navigational obstructions (such as shipwrecks, reefs, containers, and mines). Then, a "fidelity check" mechanism is introduced to ensure that points of unobstructed navigation are not mistakenly deleted by comparing the local point cloud density and spatial distribution before and after filtering.

[0006] 2) Error Correction To address the system errors caused by variations in the attitude and sound velocity profile of the multibeam system itself, a point-by-point correction algorithm based on a time-varying model is adopted. At the same time, mutual constraint adjustment is performed using the overlapping areas of adjacent strips to improve the absolute and relative accuracy of the point cloud.

[0007] 3) Adaptive thinning While maintaining the integrity of navigational obstruction features, a curvature-driven non-uniform thinning method is designed—using a larger step size for thinning in flat areas, and maintaining the original density or even locally increasing the density in areas with severe terrain undulations or suspected targets. By setting a "safety thinning threshold," it is theoretically guaranteed that the representative points of the point cloud of any potential navigational obstruction (meeting navigation safety standards at the minimum scale) will not be deleted.

[0008] 4) Rapid Identification of Underwater Navigation Obstructions: Using the previously processed high-fidelity point cloud as input, a lightweight and real-time deep learning recognition pipeline is constructed: First, target contour extraction: An improved PointNet++ or a Transformer-based point cloud segmentation network (such as Point Transformer) is used to directly perform end-to-end instance segmentation on the 3D point cloud, quickly separating underwater targets (isolated objects) from the background (seabed topography). To address the sparse and partially occluded characteristics of underwater target point clouds, multi-scale feature aggregation and edge attention mechanisms are introduced to improve the completeness and boundary accuracy of contour extraction. Second, property recognition and attribute classification: On the segmented target point cloud blocks, a point cloud convolutional network is used to extract multimodal features such as geometry, intensity, and local density. This is combined with a small classification network (such as PointCNN or dynamic graph CNN) to achieve target property recognition (such as rocks, shipwrecks, oil pipelines, containers, mines, etc.) and attribute classification (size, orientation, depth from the water surface, risk level). (3) Taking the new state information carried by the latest nautical chart as the main framework and the transient information obtained by the rapid identification of underwater navigation obstacles as the dynamic update layer, the new state and transient marine geographic information are integrated through multi-source heterogeneous data fusion, spatiotemporal consistency verification and dynamic visualization technology, and the transient information (such as transient underwater geographic information about the category, outline and attributes of underwater navigation obstacles) is accurately superimposed on the base state information.

[0009] (4) When new and transient information are fused and processed, if a new obstacle is found on the planned route that is not marked by the new information, a turning point is added at an appropriate position on the planned route, and scanning and detection are carried out from the point in the order of "first the upper left, then the upper right" or "first the upper right, then the upper left" to generate a safe obstacle bypass path. (5) After bypassing the obstacle, the vessel returns to its original planned route and continues sailing to the destination according to steps (2) to (4). Figure 2 and Figure 3 As shown in the image, the striped areas represent the coverage of transient oceanographic information, the red symbols represent navigational obstructions (reefs) identified by high-resolution multibeam echo sounders mounted on remotely operated small vessels, and the green lines represent the actual tracks of the vessels. Figure 2 The demonstration showed how a ship sails from south to north along a planned route, and when it encounters an obstacle that is detected in real time on the planned route but not marked in the new information, it detours around it in a set order and finally returns to the planned route to continue sailing north. Figure 3 This demonstrates the ship's detour around a real-time detected obstacle and its actual trajectory during the execution of the entire navigation plan. Since the obstacle (i.e., a reef not marked on the ground state information) at the intermediate position of the ship was marked on the nautical chart when the ship designed the planned route, the detour route, which is different from the ground state and transient fusion protection mode, was given in advance based on the water depth and other conditions around the obstacle.

[0010] The detection position of the remotely controlled small unmanned vessel is located in front of the ship, and the distance is set according to the ship's turning radius, with a setting of over 1000 meters for large ships. Assuming the ship is a 100,000-ton vessel, approximately 250 meters long, and has a speed of 20 knots, under full load, its turning radius for a full rudder turn is calculated to be approximately 625 meters based on 2.5 times the ship's length. Therefore, 1000 meters is taken as the minimum distance to fully meet the ship's safety distance requirements for turning and obstacle avoidance maneuvers in the waters ahead of the bow.

[0011] The high-resolution multibeam detection device has a smaller beam angle and higher resolution than conventional multibeam equipment, and supports higher ship speeds and longer continuous detection time.

[0012] Compared with the prior art, the beneficial effects of the present invention are: Integrating transient information into new products can further enhance navigation safety in complex sea areas, serving as a powerful supplement to legally mandated protection products. By using remotely controlled unmanned vessels equipped with detection devices, the limitations of directly installing detection equipment on ships in terms of detection distance, beam angle, and resolution are overcome. This provides systematic technical support and data assurance for ships navigating in complex environments such as unfamiliar waters and waters affected by extreme factors. Attached Figure Description

[0013] Figure 1 It has been updated to the latest nautical chart.

[0014] Figure 2 It is a navigation chart for ships traveling from south to north along the planned route.

[0015] Figure 3 It is a diagram showing how a ship navigates around a real-time obstruction and its actual trajectory during the execution of its entire navigation plan.

[0016] In the diagram: red crosses indicate navigational obstructions not marked in the new information; yellow solid lines represent planned routes; and green solid lines represent the actual tracks of vessels. Detailed Implementation

[0017] Based on the latest updated nautical chart information, the vessel sailed from south to north along its planned route. During the detection process, a new obstacle not marked in the updated information was discovered. The system generated a detour path in the order of "first the upper left, then the upper right," and the vessel successfully detoured and returned to its original route. This method is particularly suitable for volatile waters affected by extreme factors such as war, earthquakes, and typhoons.

Claims

1. A method for ensuring the fusion of novel and transient marine geographic information for ship navigation, comprising the following steps: (1) Based on the new marine geographic information covering the entire navigation area, which is derived from the latest paper charts or electronic nautical charts (ENC), the planned route is set; (2) The ship sails along the planned route, and a remotely controlled small unmanned vessel equipped with a high-resolution multibeam detection device conducts real-time scanning and detection along the planned route in front of the ship to obtain the transient marine geographic information required to achieve integrated support. To obtain transient information that meets the requirements of fusion processing, the raw detection data needs to undergo noise reduction, filtering, correction, thinning, and rapid identification of underwater navigational obstructions; the details are as follows: 1) Noise Reduction and Filtering First, by combining statistical outlier removal, radius filtering, and adaptive filtering based on local terrain features, noise such as suspended solids in water and detected anomalies is removed while strictly preserving the sparse multibeam point cloud corresponding to isolated navigational obstructions. Then, a "fidelity check" mechanism is introduced to ensure that unobstructed navigational points are not mistakenly deleted by comparing the local point cloud density and spatial distribution before and after filtering. 2) Error Correction To address the system errors caused by variations in the attitude and sound velocity profile of the multibeam system itself, a point-by-point correction algorithm based on a time-varying model is adopted. At the same time, mutual constraint adjustment is performed using the overlapping areas of adjacent strips to improve the absolute and relative accuracy of the point cloud. 3) Adaptive thinning While maintaining the integrity of the features of navigational obstructions, a curvature-driven non-uniform thinning method is designed—a larger step size is used for thinning in flat areas, while the original density or even local densification is maintained in areas with severe terrain undulations or suspected targets; by setting a "safety thinning threshold", the point cloud representative points of any potential navigational obstruction are theoretically guaranteed not to be deleted. 4) Rapid underwater navigation obstruction target identification: Using the previously processed high-fidelity point cloud as input, a lightweight and real-time deep learning identification pipeline is constructed: First, target contour extraction: An improved PointNet++ or a Transformer-based point cloud segmentation network (such as Point Transformer) is used to directly perform end-to-end instance segmentation on the 3D point cloud, quickly separating underwater targets (isolated objects) from the background (seabed topography). In view of the sparse and partially occluded characteristics of underwater target point clouds, multi-scale feature aggregation and edge attention mechanisms are introduced to improve the completeness and boundary accuracy of contour extraction; Second, property recognition and attribute classification: On the segmented target point cloud blocks, a point cloud convolutional network is used to extract multimodal features such as geometry, intensity, and local density, and a small classification network is combined to realize target property recognition and attribute classification. (3) Taking the new information carried by the latest nautical chart as the main framework and the transient information obtained by the rapid identification of underwater navigational obstruction targets as the dynamic update layer, the new and transient marine geographic information is integrated and processed through multi-source heterogeneous data fusion, spatiotemporal consistency verification and dynamic visualization technology, and the transient information is accurately superimposed on the new information. (4) When new and transient information are fused and processed, if a new obstacle is found on the planned route that is not marked by the new information, a turning point is added at an appropriate position on the planned route, and scanning and detection are carried out from the point in the order of "first the upper left, then the upper right" or "first the upper right, then the upper left" to generate a safe obstacle bypass path; (5) After bypassing the obstacle, the ship returns to the original planned route and continues to sail to the destination according to steps (2) to (4).

2. The method for ensuring the fusion of new and transient marine geographic information for ship navigation according to claim 1, further characterized in that: The detection position of the remotely controlled small unmanned boat is located in front of the ship, and the distance is set according to the turning radius of the ship. For large ships, it is set to more than 1,000 meters. Assuming that the ship is 100,000 tons, about 250 meters long, and has a speed of 20 knots, when it is fully loaded, the turning radius of its full-load turn is calculated to be about 625 meters based on 2.5 times the ship length. Therefore, 1,000 meters is taken as the minimum distance to fully meet the safety distance of the ship in front of the bow when turning and avoiding obstacles.

3. The method for ensuring the fusion of new and transient marine geographic information for ship navigation according to claim 1, further characterized in that: The high-resolution multibeam detection device has a smaller beam angle and higher resolution than conventional multibeam equipment, and supports higher ship speeds and longer continuous detection time.

Images