A large ship motion estimation method and system based on millimeter wave radar

Abstract

This invention discloses a method and system for motion estimation of large ships based on multi-millimeter-wave radar. The method includes: deploying at least three millimeter-wave radars along the length of the ship on both port and starboard sides to acquire radar extrinsic parameters; simultaneously acquiring point cloud data and converting it to the ship's coordinate system; filtering out isolated noise points through point cloud consistency matching in overlapping areas to obtain a clean point cloud; solving for the local velocities of each radar based on a radial velocity measurement model and the RANSAC algorithm and converting it to the ship's coordinate system; combining a rigid body kinematics model to fit and solve for the lateral / longitudinal velocities and angular velocities of the ship's center of mass, calculating the lateral / longitudinal accelerations and bow-stern lateral drift velocities, and improving stability through filtering. This invention achieves all-weather, high-precision motion state perception with no cumulative errors through multi-radar data fusion and robust algorithms, providing reliable data support for intelligent operations such as autonomous navigation and collision avoidance for large ships.

Description

Technical Field

[0001] This invention relates to the field of ship technology, and specifically to a method and system for estimating the motion of large ships based on multi-millimeter-wave radar. Background Technology

[0002] With the intelligent development of the shipping industry, accurate motion estimation of large ships has become a key technology for achieving autonomous navigation and improving safety. Due to their enormous size and inertia, large ships (such as cargo ships, tankers, and cruise ships) require extremely high precision in sensing their own motion state (including lateral and longitudinal velocities, acceleration, and angular velocities) when operating in complex environments such as ports and narrow waterways. Precise motion parameters are the foundation for path planning, dynamic collision avoidance, and automatic control systems.

[0003] Currently, ship motion estimation primarily relies on a combination of Global Positioning System (GNSS) and Inertial Measurement Unit (IMU). GPS provides global position information, but its update frequency is low, and the signal may be lost or weakened in obstructed environments such as ports, leading to decreased positioning accuracy. IMU can provide high-frequency angular velocity and acceleration information, but it suffers from drift errors that accumulate over time, requiring correction from external information (such as GNSS). Therefore, navigation systems that solely rely on a combination of GPS and IMU still have limitations in terms of reliability and accuracy.

[0004] Millimeter-wave radar, as an active sensor, boasts advantages such as all-weather operation, high resolution, high accuracy, and insensitivity to environmental changes, making it a crucial tool for positioning in situations where navigation sensors fail. By measuring the relative motion of a ship with its surroundings (such as docks, embankments, and other vessels), the ship's own motion state can be deduced. However, applying millimeter-wave radar to motion estimation for large ships faces unique challenges: 1. Significant multipath effect: Large ships are huge metal structures, and multipath noise (clutter point cloud) is easily generated around the ship, which interferes with the detection of real targets.

[0005] 2. Limitations of a single sensor: A single radar has a limited field of view and cannot cover the entire environment around the ship. It is also prone to deviations in overall motion estimation due to local measurement errors.

[0006] Therefore, there is an urgent need for a method that can effectively suppress multipath interference, fuse multiple radar information and solve the rigid body motion parameters of the ship, so as to achieve robust and real-time estimation of the motion state of large ships. Summary of the Invention

[0007] To achieve the above objectives, this invention discloses a method and system for estimating the motion of large ships based on millimeter-wave radar. By rationally deploying radar, unifying the coordinate system, and completing point cloud consistency matching and denoising, combined with the radial velocity measurement model and rigid body kinematics principles, and robustly solving the problem using the RANSAC algorithm, the method accurately estimates the ship's center of mass velocity, acceleration, and bow and stern drift velocity, overcoming multipath interference and the limitations of single sensors, and providing reliable technical support for intelligent ship navigation.

[0008] To achieve the above objectives, the present invention provides a method for estimating the motion of large ships based on millimeter-wave radar, comprising the following steps: S1. Deploy multiple millimeter-wave radars on the hull and acquire the extrinsic parameters of each radar in the hull coordinate system, the extrinsic parameters including installation position and attitude; S2. Synchronously collect point cloud data from each radar and use the external parameters to convert all point cloud data to the ship coordinate system; for any two radar pairs with overlapping fields of view, perform point cloud consistency matching in their overlapping detection areas, filter out isolated point clouds that have not formed a match, and obtain a clean point cloud for each radar. S3. For each radar, based on its clean point cloud and radar radial velocity measurement model, solve for the radar's velocity in its own coordinate system, and transform it to the ship's coordinate system to obtain the local velocity vector of each radar installation point in the ship's coordinate system. S4. Based on the rigid body kinematics model, using the local velocity vectors of all radar mounting points and their known positions in the ship's coordinate system, the RANSAC algorithm is used again to fit and solve the lateral velocity, longitudinal velocity, and angular velocity of the ship's center of mass; and the ship's lateral and longitudinal accelerations, as well as the lateral drift velocities of the bow and stern are calculated.

[0009] Furthermore, the radar deployment method is as follows: at least three millimeter-wave radars are deployed along the length of the ship on each of the port and starboard sides, located at the front, middle and rear of the hull respectively, and the detection fields of adjacent radars have overlapping areas.

[0010] Furthermore, the step of using the extrinsic parameters to transform all point cloud data into the ship's coordinate system specifically involves: The external parameters include the radar's position vector relative to the ship's coordinate system. and rotation matrix For each point cloud point detected by radar i Perform coordinate transformation: ; in, These are the coordinates of the point in the ship's coordinate system.

[0011] Furthermore, for any two radar pairs with overlapping fields of view, point cloud consistency matching is performed within their overlapping detection areas, specifically as follows: Based on the extrinsic parameters of radars R1 and R2 and their field of view (FOV), the spatial intersection region of the two radar fields of view is calculated in the ship's coordinate system. ; From the point cloud set of radar R1 Select the subset of point cloud located within the overlapping region. Similarly, from the point cloud set of radar R2 Selecting a subset of point clouds .

[0012] For point cloud subsets Each point in In the point cloud subset Search for the point that is spatially closest to it. If the Euclidean distance between two points satisfies: ; in If the distance threshold is preset, then the point is considered a pair. They are consistent, meaning they correspond to the same real target point; Point clouds that have matching points found in both radars are marked as consistent point clouds; point clouds that appear only in radar R1 or radar R2 and cannot have matching points are marked as isolated point clouds.

[0013] Furthermore, step S3 specifically includes: Based on the clean point cloud, an overdetermined set of equations is constructed. After removing potential dynamic target points or residual noise points through the RANSAC algorithm, the two-dimensional plane velocity of the radar relative to its own coordinate system is solved by the least squares method on the inner point set of RANSAC, combined with the radar radial velocity measurement model. Repeat the above calculation for all N radars to obtain the two-dimensional plane velocity of each radar relative to its own coordinate system. ; Convert the two-dimensional plane velocity of each radar relative to its own coordinate system to its velocity relative to the ship's coordinate system, thus obtaining the local velocities of the ship at N different locations in the ship's coordinate system. .

[0014] Furthermore, the radar radial velocity measurement model is as follows: ; in, It is the radial velocity measured by radar. It is the current radar's vector velocity in its own coordinate system. It is the unit vector of the direction of the target point.

[0015] Furthermore, the rigid body kinematic model is as follows: ; in, It is the position vector of radar i relative to the ship's center of mass. Let be the velocity of any point on the hull in the hull coordinate system. Let the velocity of the ship's center of mass in the ship's coordinate system be... ; On a two-dimensional plane, Expanded to: ; ; in, These are the transverse and longitudinal velocities of the ship's center of mass in the ship's coordinate system. It is the angular velocity of the ship's hull. These are the position coordinates of radar i in the ship's coordinate system.

[0016] Furthermore, step S4, calculating the ship's acceleration, specifically involves: ; ; in, Let be the lateral acceleration of the ship at time t. Let be the longitudinal acceleration of the ship at time t. Let t be the lateral velocity of the ship's center of mass. Let be the longitudinal velocity of the ship's center of mass at time t; The lateral / longitudinal velocities, lateral / longitudinal accelerations, and angular velocities of multiple consecutive frames are filtered using a Kalman filter or a moving average filter to obtain the filtered lateral velocities of the ship. Longitudinal velocity lateral acceleration Longitudinal acceleration angular velocity .

[0017] Furthermore, step S4 calculates the lateral drift speeds of the ship's bow and stern, specifically as follows: In the ship's coordinate system, let the center of mass of the ship be the origin O, the longitudinal axis y-axis point towards the bow, and the transverse axis x-axis point towards the starboard side of the ship. The position vectors of the bow and stern points in the ship's coordinate system are as follows: ; ; in, It is the longitudinal distance from the bow to the center of mass. It is the longitudinal distance from the stern to the center of mass; According to the rigid body motion relationship, in the ship's coordinate system, the lateral drift velocity is the velocity component in the x-direction: ; ; The lateral drift speed of the bow. The lateral drift speed of the stern. The filtered lateral velocity of the ship's center of mass. ω is the angular velocity of the ship.

[0018] On the other hand, this invention proposes a system for implementing the above-mentioned method for estimating the motion of large ships based on multi-millimeter-wave radar, comprising: The radar extrinsic parameter acquisition module is used to acquire the extrinsic parameters of each radar in the ship's coordinate system, including the installation position and attitude. The point cloud data acquisition and processing module is used to synchronously acquire point cloud data from each radar and use the extrinsic parameters to convert all point cloud data to the ship coordinate system. For any two radar pairs with overlapping fields of view, point cloud consistency matching is performed in their overlapping detection areas to filter out isolated point clouds that have not formed a match, thus obtaining a clean point cloud for each radar. The local velocity calculation module, based on its clean point cloud and radar radial velocity measurement model, solves the radar's velocity in its own coordinate system and transforms it to the ship's coordinate system to obtain the local velocity vector of each radar installation point in the ship's coordinate system. The motion calculation module uses the local velocity vectors of all radar mounting points and their known positions in the ship's coordinate system to again use the RANSAC algorithm to fit and solve the lateral velocity, longitudinal velocity, and angular velocity of the ship's center of mass; and calculates the ship's lateral and longitudinal accelerations, as well as the lateral drift velocities of the bow and stern.

[0019] The beneficial effects of the above-described technical solution of the present invention are as follows: 1. This invention effectively overcomes the limitations of traditional ship motion estimation techniques through an innovative multi-millimeter-wave radar deployment and data fusion scheme. The design of at least three radars deployed aft, midway, and forward on each side, combined with overlapping field-of-view planning, achieves comprehensive coverage of the surrounding environment and accurately filters out clutter point clouds caused by multipath effects through a point cloud consistency matching mechanism, solving the detection interference problem caused by large metal hulls. Combining the two-round robust solution of the RANSAC algorithm not only eliminates the influence of dynamic targets and noise but also significantly improves the estimation accuracy of core parameters such as center-of-mass velocity and angular velocity by fusing local velocity data from multiple radars through a rigid body kinematics model. Compared to the GNSS and IMU combination scheme, this method does not rely on satellite signals, is unaffected by port obstruction, and has no cumulative drift error, achieving all-weather, high-real-time motion state perception.

[0020] 2. This invention provides key technical support for the intelligent operation of large ships, significantly improving shipping safety and operational efficiency. The precisely output parameters such as lateral / longitudinal velocity, acceleration, and bow-stern lateral drift velocity are adaptable to complex scenarios such as port berthing and navigation in narrow waterways, providing a reliable data foundation for path planning, dynamic collision avoidance, and automatic control systems. Attached Figure Description

[0021] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a schematic diagram of the millimeter-wave radar deployment according to the present invention; Figure 3 This is a diagram showing the system module composition of the present invention. Detailed Implementation

[0022] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0023] like Figure 1 As shown, the present invention provides a method for estimating the motion of a large ship based on millimeter-wave radar, comprising the following steps: S1. Deploy multiple millimeter-wave radars on the hull and acquire the extrinsic parameters of each radar in the hull coordinate system, the extrinsic parameters including installation position and attitude; S2. Synchronously collect point cloud data from each radar and use the external parameters to convert all point cloud data to the ship coordinate system; for any two radar pairs with overlapping fields of view, perform point cloud consistency matching in their overlapping detection areas, filter out isolated point clouds that have not formed a match, and obtain a clean point cloud for each radar. S3. For each radar, based on its clean point cloud and radar radial velocity measurement model, solve for the radar's velocity in its own coordinate system, and transform it to the ship's coordinate system to obtain the local velocity vector of each radar installation point in the ship's coordinate system. S4. Based on the rigid body kinematics model, using the local velocity vectors of all radar mounting points and their known positions in the ship's coordinate system, the RANSAC algorithm is used again to fit and solve the lateral velocity, longitudinal velocity, and angular velocity of the ship's center of mass; and the ship's lateral and longitudinal accelerations, as well as the lateral drift velocities of the bow and stern are calculated.

[0024] The following details the implementation of the above method: In step S1, the installation location of the millimeter-wave radar is first rationally planned. Based on the ship's design drawings, the radar's installation location and orientation are determined. To ensure measurement coverage and data redundancy, the deployment principles are as follows: The ship is equipped with at least three millimeter-wave radars on each side, located at the front, middle and rear of the hull respectively.

[0025] To achieve multipath suppression, there should be a clear overlap of field of view between adjacent radars.

[0026] Ideally, one radar would be installed directly forward of the bow, and one each on the port and starboard sides; along the sides of the ship, radars would be installed at intervals of approximately 30 meters, depending on the length of the hull.

[0027] The radar is installed horizontally, and the area of ​​the ship itself within the radar's field of view is minimized as much as possible.

[0028] After installation, the position and attitude parameters (i.e., extrinsic parameters) of each radar in the ship's coordinate system are accurately measured. These extrinsic parameters are the basis for subsequent motion model calculations. Figure 2 This paper presents a typical radar deployment scheme, showing example installation locations of the front, left, and right radars and their approximate detection ranges.

[0029] Step S2 mainly achieves multipath effect suppression based on multi-radar cross-verification, specifically including: (1) Unification of coordinate system The point cloud data output by each millimeter-wave radar is typically located in its own sensor coordinate system, with the radar's installation location as the origin and the orientation as the coordinate axis. To achieve point cloud comparison and verification between multiple radars, it is first necessary to unify the point cloud data of all radars to the same reference coordinate system. This invention uses the ship's hull coordinate system as the unified reference system.

[0030] Specifically, the radar extrinsic parameters obtained in step one (including the radar's position vector relative to the ship's coordinate system) are used. and rotation matrix For each point cloud point detected by radar i Perform coordinate transformation: ; in, This refers to the coordinates of the point in the ship's coordinate system. Through this transformation, the point cloud data of all radars are mapped to a unified ship's coordinate system, laying the foundation for subsequent spatial comparison.

[0031] (2) Selection of overlapping regions For any two radars with overlapping fields of view (e.g., radar R1 and radar R2), it is necessary to accurately determine their high-confidence overlapping detection area. This area is selected based on the following principles: Field of view intersection: Based on the extrinsic parameters (position and orientation) of radars R1 and R2 and their field of view (FOV), calculate the spatial intersection area of ​​the two radar fields of view in the ship coordinate system.

[0032] Detection range constraint: Considering the effective detection range and signal-to-noise ratio characteristics of the radars, only areas within the effective detection range of both radars (typically 5 to 80 meters) are selected to ensure point cloud quality, resulting in the effective overlap area. .

[0033] (3) Point cloud consistency judgment After determining the overlapping area Then, the consistency of the point cloud located in this area is checked. The basic principle is that a real physical target point (such as the edge of a dock, a fixed obstacle, etc.) should theoretically be detected by both radar R1 and radar R2 at the same time and have similar spatial positions in the ship's coordinate system.

[0034] The specific judgment steps are as follows: Point cloud filtering: from the point cloud set of radar R1 Select the subset of point cloud located within the overlapping region. Similarly, from the point cloud set of radar R2 Selected from .

[0035] Spatial matching: For Each point in ,exist Search for the point that is spatially closest to it. If the Euclidean distance between two points satisfies: ; in If the distance threshold is preset (usually 1-2 meters, determined according to radar resolution), then the point is considered a match. They are consistent, meaning they correspond to the same real target point.

[0036] Consistency labeling: Point clouds that have matching points found in both radars are labeled as "consistent point clouds"; point clouds that appear only in radar R1 or radar R2 and cannot have matching points are labeled as "isolated point clouds".

[0037] (4) Impurity filtration Based on the point cloud consistency assessment, multipath noise filtering is performed, deleting "isolated point clouds" and retaining "consistent point clouds." This verification and filtering process is applied to all radar pairs with overlapping fields of view (e.g., R1-R2, R2-R3, R3-R4, etc.). Through multiple iterations, multipath noise is gradually eliminated from each radar point cloud, ultimately yielding multi-radar point cloud data after global multipath filtering. .

[0038] Step S3 achieves single-point velocity estimation of the ship based on a single millimeter-wave radar, specifically as follows: 1. Velocity Estimation Model: Millimeter-wave radar can directly measure the relative radial velocity of a target point. For a stationary background environment (such as a dock), the velocity of the point cloud detected by the radar reflects the motion of the ship itself. As a rigid body, the velocity of any point on the ship conforms to the laws of rigid body motion. Assume that within a short time, the radial velocities of multiple stationary background points detected by the radar have the following relationship with the radar's own velocity: ; in, It is the radial velocity measured by radar. It is the current radar's vector velocity in its own coordinate system. It is the unit vector of the direction of the target point.

[0039] 2. Velocity Calculation: For a single radar, an overdetermined system of equations is constructed using all valid point clouds detected by it after verification in step two. To enhance the robustness of the results, the Random Sample Consensus (RANSAC) algorithm is used to eliminate potential dynamic target points (such as other moving vessels) or residual noise points. On the interior point set of RANSAC, the two-dimensional planar velocity of the radar relative to its own coordinate system is solved using the least squares method.

[0040] 3. Multi-point velocity acquisition: Repeat the above calculation for all N radars to obtain the two-dimensional plane velocity of each radar relative to its own coordinate system. .

[0041] 4. Velocity Coordinate System Transformation: Based on the millimeter-wave radar installation angle, the velocities of multiple radars relative to their own coordinate system are transformed into velocities relative to the ship's hull coordinate system, resulting in the local velocities of the ship at N different locations (in the ship's coordinate system). Down) .

[0042] Step S4 implements the overall motion estimation of the ship based on multi-point velocities. The implementation steps include: Rigid body motion model: the velocity of any point on the hull (i.e., the radar mounting point). (Velocity in the ship's coordinate system) and the velocity of the ship's center of mass (in the ship's coordinate system) and the ship's angular velocity The relationship between them is as follows: ; in, This is the position vector of radar i relative to the ship's center of mass (obtainable from the extrinsic parameters in step one). In a two-dimensional plane, this formula can be expanded as: ; ; here, These are the transverse and longitudinal velocities of the ship's center of mass (in the ship's coordinate system). It is the angular velocity of the ship's hull. These are the position coordinates of radar i in the ship's coordinate system.

[0043] Overall motion fitting: using the N local velocities obtained in step three and known radar position vector Construct a system for unknown quantities The overdetermined linear equations were fitted. The RANSAC algorithm was also used for fitting to avoid the influence of erroneous velocity estimates from individual radars on the final results. RANSAC can effectively identify and eliminate radar data that does not conform to the overall motion model, thus obtaining robust lateral, longitudinal, and angular velocities of the hull.

[0044] Acceleration calculation and smoothing: By differentiating the velocity estimation results of consecutive time frames, the lateral and longitudinal accelerations of the ship can be calculated.

[0045] ; ; To improve the smoothness and stability of the output, the lateral / vertical velocities, lateral / vertical accelerations, and angular velocities of multiple consecutive frames are filtered (e.g., using a Kalman filter or a moving average filter) to obtain the filtered lateral velocities of the ship. Longitudinal velocity lateral acceleration Longitudinal acceleration angular velocity .

[0046] Bow and stern velocity calculation: Based on the estimated center of mass velocity and angular velocity, the velocities of key points on the ship (such as the bow and stern) can be further calculated as auxiliary information output.

[0047] In the ship's coordinate system, let the center of mass of the ship be the origin O, the longitudinal axis (y-axis) point towards the bow, and the transverse axis (x-axis) point towards the starboard side of the ship. The position vectors of the bow and stern points in the ship's coordinate system are as follows: ; ; in, It is the longitudinal distance from the bow to the center of mass. It is the longitudinal distance from the stern to the center of mass (all positive values).

[0048] For ship operators, the more important factors are the lateral drift velocities at the bow and stern, i.e., the velocity components perpendicular to the ship's longitudinal axis. According to rigid body motion principles, in the ship's coordinate system, the lateral drift velocity is the velocity component in the x-direction: ; ; Ship information output: The final output information includes the smoothed lateral and longitudinal velocities, angular velocities, and lateral velocity of the ship. Longitudinal velocity lateral acceleration Longitudinal acceleration angular velocity And the calculated lateral drift speeds of the bow and stern .

[0049] like Figure 3 As shown, this invention proposes a system for implementing the above-mentioned method for estimating the motion of large ships based on multi-millimeter-wave radar, comprising: The radar extrinsic parameter acquisition module 101 is used to acquire the extrinsic parameters of each radar in the ship's coordinate system, including the installation position and attitude. The point cloud data acquisition and processing module 102 is used to synchronously acquire point cloud data from each radar and use the external parameters to convert all point cloud data to the ship coordinate system. For any two radar pairs with overlapping fields of view, point cloud consistency matching is performed in their overlapping detection areas to filter out isolated point clouds that have not formed a match, thus obtaining a clean point cloud for each radar. The local velocity calculation module 103, based on its clean point cloud and radar radial velocity measurement model, solves the velocity of the radar in its own coordinate system and transforms it to the ship's coordinate system to obtain the local velocity vector of each radar installation point in the ship's coordinate system. The motion calculation module 104 uses the local velocity vectors of all radar mounting points and their known positions in the ship's coordinate system to again use the RANSAC algorithm to fit and solve the lateral velocity, longitudinal velocity, and angular velocity of the ship's center of mass; and calculates the ship's lateral and longitudinal accelerations, as well as the lateral drift velocities of the bow and stern.

[0050] In summary, this invention provides a robust and efficient motion estimation scheme for large ships. Its core breakthrough lies in overcoming the limitations of traditional navigation schemes through the fusion of multi-millimeter-wave radar collaboration and precise algorithms. It innovatively employs a layout with at least three radars deployed along the ship's length on both the port and starboard sides, achieving full coverage of the surrounding environment through overlapping fields of view. Combined with a point cloud consistency matching mechanism, it accurately filters out isolated noise points caused by multipath effects, obtaining a highly reliable and clean point cloud. Relying on a radial velocity measurement model and the RANSAC algorithm for two rounds of robust solution, the local velocities at each radar installation point are first obtained. Then, through a rigid body kinematics model, multi-source data is fused to accurately calculate the lateral / longitudinal velocities and angular velocities of the center of mass. Further, acceleration and bow-stern lateral drift velocities are derived, and filtering enhances output stability. This scheme does not rely on satellite signals, is unaffected by obstructed environments, and has no cumulative errors, achieving all-weather high-precision motion perception. It provides comprehensive and reliable data support for intelligent operations such as autonomous navigation and collision avoidance for large ships, possessing significant engineering application value and promising prospects for widespread adoption.

[0051] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for estimating the motion of large ships based on multi-millimeter-wave radar, characterized in that, Includes the following steps: S1. Deploy multiple millimeter-wave radars on the hull and acquire the external parameters of each radar in the hull coordinate system, the external parameters including installation position and attitude; S2. Synchronously collect point cloud data from each radar and use the external parameters to convert all point cloud data to the ship coordinate system; for any two radar pairs with overlapping fields of view, perform point cloud consistency matching in their overlapping detection areas, filter out isolated point clouds that have not formed a match, and obtain a clean point cloud for each radar. S3. For each radar, based on its clean point cloud and radar radial velocity measurement model, solve for the radar's velocity in its own coordinate system, and transform it to the ship's coordinate system to obtain the local velocity vector of each radar installation point in the ship's coordinate system. S4. Based on the rigid body kinematics model, using the local velocity vectors of all radar mounting points and their known positions in the ship's coordinate system, the RANSAC algorithm is used again to fit and solve the lateral velocity, longitudinal velocity, and angular velocity of the ship's center of mass; and the ship's lateral and longitudinal accelerations, as well as the lateral drift velocities of the bow and stern are calculated.

2. The method for estimating the motion of large ships based on multi-millimeter-wave radar according to claim 1, characterized in that, The radar deployment method is as follows: at least three millimeter-wave radars are deployed along the length of the ship on each of the port and starboard sides, located at the front, middle and rear of the hull respectively, and the detection fields of adjacent radars have overlapping areas.

3. The method for estimating the motion of large ships based on multi-millimeter-wave radar according to claim 1, characterized in that, The specific steps for converting all point cloud data to the ship's coordinate system using the extrinsic parameters are as follows: The external parameters include the radar's position vector relative to the ship's coordinate system. and rotation matrix For each point cloud point detected by radar i Perform coordinate transformation: ; in, These are the coordinates of the point in the ship's coordinate system.

4. The method for estimating the motion of large ships based on multi-millimeter-wave radar according to claim 1, characterized in that, The step of performing point cloud consistency matching within the overlapping detection area for any two radar pairs with overlapping fields of view includes the following steps: Based on the extrinsic parameters of radars R1 and R2 and their field of view (FOV), the spatial intersection region of the two radar fields of view is calculated in the ship's coordinate system. ; From the point cloud set of radar R1 Selecting subsets of point clouds located within overlapping regions Similarly, from the point cloud set of radar R2 Selecting a subset of point clouds For point cloud subsets Each point in In the point cloud subset Search for the point that is spatially closest to it. If the Euclidean distance between two points satisfies: ; in If the distance threshold is preset, then the point is considered a pair. They are consistent, meaning they correspond to the same real target point; Point clouds that have matching points found in both radars are marked as consistent point clouds; point clouds that appear only in radar R1 or radar R2 and cannot have matching points are marked as isolated point clouds.

5. The method for estimating the motion of large ships based on multi-millimeter-wave radar according to claim 1, characterized in that, Step S3 specifically involves: Based on the clean point cloud, an overdetermined set of equations is constructed. After removing potential dynamic target points or residual noise points through the RANSAC algorithm, the two-dimensional plane velocity of the radar relative to its own coordinate system is solved by the least squares method on the inner point set of RANSAC, combined with the radar radial velocity measurement model. Repeat the above calculation for all N radars to obtain the two-dimensional plane velocity of each radar relative to its own coordinate system. ; Convert the two-dimensional plane velocity of each radar relative to its own coordinate system to its velocity relative to the ship's coordinate system, thus obtaining the local velocities of the ship at N different locations in the ship's coordinate system. .

6. The method for estimating the motion of large ships based on multi-millimeter-wave radar according to claim 5, characterized in that, The radar radial velocity measurement model is as follows: ; in, It is the radial velocity measured by radar. It is the current radar's vector velocity in its own coordinate system. It is the unit vector of the direction of the target point.

7. The method for estimating the motion of large ships based on multi-millimeter-wave radar according to claim 1, characterized in that, The rigid body kinematics model is as follows: ; in, It is the position vector of radar i relative to the ship's center of mass. Let be the velocity of any point on the hull in the hull coordinate system. Let the velocity of the ship's center of mass in the ship's coordinate system be... ; On a two-dimensional plane, Expanded to: ; ; in, These are the transverse and longitudinal velocities of the ship's center of mass in the ship's coordinate system. It is the angular velocity of the ship's hull. These are the position coordinates of radar i in the ship's coordinate system. Let i be the lateral velocity component of the radar mounting point in the ship's coordinate system. Let be the longitudinal velocity component of radar i's mounting point in the ship's coordinate system.

8. The method for estimating the motion of large ships based on multi-millimeter-wave radar according to claim 1, characterized in that, Step S4, which calculates the ship's acceleration, specifically involves: ; ; in, Let be the lateral acceleration of the ship at time t. Let be the longitudinal acceleration of the ship at time t. Let be the lateral velocity of the ship's center of mass at time t. Let be the longitudinal velocity of the ship's center of mass at time t; The lateral / longitudinal velocities, lateral / longitudinal accelerations, and angular velocities of multiple consecutive frames are filtered using a Kalman filter or a moving average filter to obtain the filtered lateral velocities of the ship. Longitudinal velocity lateral acceleration Longitudinal acceleration angular velocity .

9. The method for estimating the motion of large ships based on multi-millimeter-wave radar according to claim 1, characterized in that, Step S4, which calculates the lateral drift speeds of the ship's bow and stern, specifically involves: In the ship's coordinate system, let the center of mass of the ship be the origin O, the longitudinal axis y-axis point towards the bow, and the transverse axis x-axis point towards the starboard side of the ship. The position vectors of the bow and stern points in the ship's coordinate system are as follows: ; ; in, It is the longitudinal distance from the bow to the center of mass. It is the longitudinal distance from the stern to the center of mass; According to the rigid body motion relationship, in the ship's coordinate system, the lateral drift velocity is the velocity component in the x-direction: ; ; The lateral drift speed of the bow. The lateral drift speed of the stern. The filtered lateral velocity of the ship's center of mass. ω is the angular velocity of the ship.

10. A system for implementing the motion estimation method for large ships based on multi-millimeter-wave radar as described in any one of claims 1-9, characterized in that, include: The radar extrinsic parameter acquisition module is used to acquire the extrinsic parameters of each radar in the ship's coordinate system, including the installation position and attitude. The point cloud data acquisition and processing module is used to synchronously acquire point cloud data from each radar and use the extrinsic parameters to convert all point cloud data to the ship coordinate system. For any two radar pairs with overlapping fields of view, point cloud consistency matching is performed in their overlapping detection areas to filter out isolated point clouds that have not formed a match, thus obtaining a clean point cloud for each radar. The local velocity calculation module, based on its clean point cloud and radar radial velocity measurement model, solves the radar's velocity in its own coordinate system and transforms it to the ship's coordinate system to obtain the local velocity vector of each radar installation point in the ship's coordinate system. The motion calculation module uses the local velocity vectors of all radar mounting points and their known positions in the ship's coordinate system to again use the RANSAC algorithm to fit and solve the lateral velocity, longitudinal velocity, and angular velocity of the ship's center of mass; and calculates the ship's lateral and longitudinal accelerations, as well as the lateral drift velocities of the bow and stern.

Images